E’ERCEP'EEGN G? AWLITUEE WfiPEESSED
SFEECB BY FEESONS EXEEEETEHG
LO‘UB'NESS EECHETMR?

Thais gar “to Degree a? pit. D.
MECEESAN STATE UREVEBSETY
Samuei Brit‘co‘a Burchféeicfi
W70

   
 

  

LIBRARY

Michigan State
University

{fit-.3913

    

This is to certify that the

thesis entitled

PERCEPTION OF AMPLITUDE COMPRESSED
SPEECH BY PERSONS EXHIBITING
LOUDNESS RECRUITMENT
presented by

Samuel Britton Burchfield

has been accepted towards fulfillment
of the requirements for

Ph 0 Do degree in Audi 01 083'

. i/7‘ _
/Mi’:’”mb /L»5 7

'1 ’1“ ”WV
Major pr¢fessor

Due 11-3-1970

0-169

Samuel Britten Burchfield

Sseech discrimination Derformance of thirty-six
subjects was compared for conditions of amplitude
comnression and for no compression. These conditions
were achieved with a nrototrne amelitude comnressor. The
instrument had innut-to—outnut amplification ratios of
two~to-one and three—to—one and it also functioned as a
linear amplifier which was called one-to-one annlification.

The thirty—six subjects, all of whom had unilateral
sensorineural hearinq loss, were partitioned into partial
or complete loudness recruitment brouns on the basis of
their reseonse to a sneech noise alternate binaural
loudness balanace (ABLE) test. Classification of subjects
was carried out accordinq to two procedures. The "classical"
nrocedure involved a comnarison of the maenitude of the
sensation levels above threshold necessary to make the
test sianals eouallv loud. The second system was based on

intra-aural comparison of the loudness growth function.

banuel Eritton Burchfield

Snee h discrimination tests eanloyine monosyllabic
CNC words were nresented at 2h d3 sensation level at the
affected ear under each comnression condition. Eroad band
maskins noise was routinely anblied at the contralateral
ear.

Statistical analyses of the data, partitioned by
either method for defining recruitment, yielded similar
results. Sneech discrimination scores were sienificantly
hieher for the two~to-one and the three-to-one comuression

than for the one-to-one and the benefit was OI the same

order of masnitude for both recruitment groups.

or A" ‘ “

hr

-RLSSEU

IITULE CCLI

fi‘trv' T" mI‘f'fl
a _J .
1! 31:3 [31.1. LUJ
. fins v- {‘1
A-Jb’ H

. 4

_nCRUITWEEUP

By
Samuel Britton

Burchfield

Submitted to
I
in aartial

nichisan State University
,_ fulfillment of the

no requirements
for the deeree of

DCCTCR CF FHIICSCFHY

Denartment of Audioloey and
_ Sneech Sciences

1970

CQConyriiht by
SAhUEL BRITTCN BURGER ElD

1971

ii

(-3: Com/right by
SALUEL BRITTCN BURCHFIELD

1971

ii

Please Note:

Some pages have very light
-type. Filmed as received.

University Microfilms.

my Wife Burch,

gave us Christy

iii

ACILL“ .C'w'lE C—E ELLIS

I would like to thank a number of individua s whoa
guidance and assistance enabled the coneletion of thi
study. The followina neonle are eratefully acknowledéed:

q

, o o ‘-, 0 . ‘ a, h. T- , ”T ath I. -
r. Nillian mintelmann, chai men, and Jr. RGILOll

U

Cver, Dr. lco Deal, Dr. Bernice Borcman, and Dr. Oscar
Tosi, committee members.
Lr. iatrick Carter, Dr. Sanford Snyderman, Dr. E. n.

in, and Dr. John Thomas (Ear,

I,
(3
la
L 2
3
¥
9.)
{E
o.
d
o
O
5..
0.)
LJ
H
(D
U)
(
Po
1—13
rs

Nose, and Throat Associates Inc., Fort wavne) for
(ssistance in obtaininq subjects.

Dr. Judith Fre .nkmann and Dr. Bradley Lashbrook for
their exnerimental and statistical guidance.

Kr. Richard Brander and hr. Erwin Weiss (Eeltcne
Electronics Co., Chicano) and tr. Donald Riggs for
assistance with the instrumentation emnloyed.

I would also like to extend my sincere gratitude
to Dr. Jud. ith r'anxmcnn for her interest and encouraQr-ment.

The study was sunnorted in Dart by an American
Steech and Hearind roundation Traineeshin \ hich vm s funded

/ ,V

by grant No. T—oZ-lO7 from the lublic Health Services.

1‘!

1—1

TADLE CF CCNiEjTS

(W
L I
(«—4
C)
e
H
('1
O
O
O
C
O
O
O
O
O
O
O

:eriormance Characteristics oi thlllUQ
Controllinfi Devices . . . . . . . .
In»ut-Luteut Relationshins . . .
E'iI'LIC COLESEU Its a o o a a a c o a
Zrobloms Associated Visa AuUliEIEG
COE’E-I'E-.Ol I I I I I I I I I I I I I I
111,175le 0 o o o o a o a o o I o o
.rroeier1ahans EMUl rum.dmr‘ . . . .
*Vcessive Noise Level . . . . .
Effects of Amelitude Control on “nee
Discriminati n . . . . . . . . . . .
ftnlitrfie Control as a Sol Wion f r
EOU’QOSS Recruitment . . . . . . . .
”2-1 11112.7"! 0 o o a o a o o o o o o o a
EfILEL TAl IfiCCEZdLiE” . . . . . . . .

Sinnjewztis . . . . . . . . . . . . . .
Recruitment ;uantii ic ation . . . . .
Seee,h Koise A313 Test , . , , .
lntern.etation of the Sneech Hoi
1E- .3 LEGS-I; a o a c a o o a c o a

Cal-"Q'ECII I thz’Bl‘lleF o a o o o a a o c o
lhe ”eiss AAtli ude Comnressor . . .
Innut-Cutnut Calibration . . . .

hiss-Decav Transients . . . . .
harrnonic Dis .ortion . . . . . .
E‘OELSG Lexie]... a a o o o a o o

I I
Ere enaration 03 th e CNC Test Stimuli
avnerinental lrocedures . . . . . .
1

urc Tone Threshold Tests . . .
S eech Recention ir,sholu Tests

V

«'3

‘

o (D.

(D

(i (r O
‘ k,. V

‘_.

\\4\)

Ln

' l
‘ \~P\ \J'EK

\ “E

I"
O

11

H
H

E El\))

K— N. K- N
" . -.\- ~7— -._ ‘40 LL) \A.) KJJ I‘) {‘3 {‘0
KJJ \0 CAR C\ {\‘I \‘E a

I
\ I

.'J\
L: H 1—4 um an -;‘

In
\_.
K
L

'1 _:C*)C\I 1' Jim,

.-‘
1vnrv‘r) n
bLL..; ...... y

\fx-r r‘rw'}
L-AI—‘D

14K

1... .LJ

IE]. 111:.)

Analvsis

?
.5
'J
I...)
'3
I.)
O
‘rj

Sunnlerental :n
Annlitude Comer.
Summary of
Discussion of

Does ambl i

D .33

\

fi

1
_’\
l

0

(v

0 Cr‘ (4.

3' C :7 CD
1—“ CL ('0 (D
.3 O

ears? . . .
Is the arxnm

seeech die or
‘the :8A ourrt C
re cruitr cnt?
Is the amoun
seecch discr

I A“ re
tn» denies 0
Summary and Clin

U)

Lli.P1€117)’ . .
Recruitment

1 17 Test .
7xtettuil val-

TW‘ we.
.1

V. SUPRARY, CCNCIL SICK S,
SUTWQTV . . o o

Conclusions . .
Recc m cndations

AV

...--.. I I I I I I I I

71777C-;-

Al" IE’

L-T‘, “with
-1.LJII.1JLD . e I I I I I I I

q 2* 4- .
clementati

Anln moht
.J .34 .L -1 A.) I I I I I I

1: _,_ - .
.L.‘ :3. L; I: C ’37, I I I I I I I
I I I I I I I I I I I
I if:
\. . t I I I I I I I I I I
’ T“ n .A'
tice 3_1ects . . . . .
E”Cects f AAnlituee
I I I I I I I I I I I
O .p .1- . 1‘ _fl 1“
lysis 0. use allects o
I
8-10fl . . . . . . . .

Hesults . . . . . .
sion 1m .rove
tion in re ecruitin

. . . . .
nt of nrovemrt
iAination related
f t1 e loudn-ss

. . . . . . . . . . .
t f imnrovement in
imination related to
f the comnression? . .
ical lnnlications . .

I I I I I I I I I
and the Sneech Noise

iditv

3')
l r- -‘
AJ
1...)
l-
I
11')

I
N J",-
'11.

. . . . . . .
of the Fesults .
on of Amnlitu
I I I I I I I I 0 O I

.A”D RECCTLEIfllnilCFS

H)

'1'

".

.
!
\J

O

l

.1

I
\E

\
E13 UVKR

I‘V‘n

\] \1 CM
CVJ‘. \3‘

\1
O

('1‘, C7": (‘0
Amped

V“.
\O

Table face
1. 2*e and sex distributions of the two
e::ficrj-vlenta.l (VFOLITFS I Q g o g g g I . g g g . 0 22A
2. EercentaAe distortion of the second and third
h. aIAonics of the Weiss Cor1hre ssor for one—to-
one, he -to- -one, and three-to-one
0031*“17‘8 8851.011 I I I I I I I I I I I I I I I I I I [,5
3. Lean oercent correct discrimination scores for
the Are-evner1rcnt1]_and host— —exnerinental
tests for the partial and the complete
recruitment srouns . . . . . . . . . . . . . . 61
h. Kean nerceht correct discrim1ha+10h scores
for"the tfi11rt1nusix asl1iectxs :ior"the 'first,
second, and third presentation orders for
the one-, two—, and three- to- -one corrres110d
conditions . . . . . . . . . . . . . . . . . . 62
5. Sum“ rv of analysis of covari_ance comeazin,
eeI: ormance of the two recruitment Wioues
for one-, tmo—, an; thre 03- -to- ~one “Arlitll
,,
cotili’renmlon I I I I I I I I I I I I I I I I I I CO
e. Duncan's new. Aultinle rsnse test anolied to
the c1rirrences between the three comeression
I‘a'thS (P. = )6) I I I I I I I I I I I I I I I O 67
7. Summary of unweishted—means analysis of
variance comnarins performance of Class I
and Class II recruitment for the one-, two-,
and three-to-one comeression ratios . . . . . . 7M
8. Beans, ranscs and standard deviations of
eercent Wain in eech dis scrimination scores
from the one-to— -one ratio to the two —to-one
ratio, and from the one- -to- -one ratio to the
three-to-one ratio (N = 36) . . . . . . . . . . 75

FiVure

{\J

awn oh racter1stics of 11V iter ano
comnressors. (A-E—C) linea W1 anoliiier

1:1 ratio, (A 1-D) compressor 2:1 re 10,
(A-E) comnressor 3:1 ratio, (A- B—b) 1L iter
3:1 ra tio, (P-E- 3) limiter 1:0 ratio (irom
Carawe V and Carhert,1967, o. lh25) . . . .

DVnamic r8 QVe reduction oroduced bV one-
tc-one, two- to- -one, and th1i'ee- -to- one
amnlitude comoression for sneech with a

30 d5 anc1m1-c Iane . . . . . . . . . . . .

edian nure—tone thresholds for the test

ar solid line) and the nontest ear

dashed line) for both recruitment crouns.
claritV, oan the air—conduction

est olds a~e shown. The ranxes of

esholos are shown for the test e2_ . . .

r—4

H‘ c+ “If A (D
O

0 Katie cis.Vram of the o “Ch-301s;
t-rnate Bine ural loud css Bale fince
:K'Cij.n"-r COLIiT‘uiPJYIt . O O Q Q C O O O O O .

ch snec;1_1;m of the sneech n01sc

n
J
. \' ‘» l .
L I -' ‘QC" 0'1"“ C 0‘
:1-...:..’.. “LVKJI L)” (1410 o o o o o o o o 0 o o o o

HVnothetical -xamn1es illustr¢1t1-n0 the
traditional method for intreoreting results
of the alternate binaurol loudness balance
test (0: Vooo ear; X: boor ear) . . . . . .

,. .1- ' .-.~ .17- “'4 -- -x- ‘ n '
maulC die _ ram oi the bub tesc stixali
l V V
.A

,_ 'nvwf' .r- r:
(.3.- ‘- Kill.» 2.1.19 0 o o I o o o o o o o I

l;~1out-outnut iunctions of the V'd— c-meunel
(ior a 1500 :2 siVna 1) for each corn r :cession

V111

\O

«“0

-\ J

‘11)
KJ)

A“
\

L1)

11

v-
t

L

“recrencv DDDDD D or t
Drocescln" Dvsvem “1e

‘l'Jq "'t3
u x/J” _/

. \ _‘ 4 q ",

LTD-Lo-o e A , and bfl ee

. ~ -u w. u m “. ‘3 '.. I .1. .

.eaD CDDGCD leCIImlDaLlOH
' -\ \ '1 " f ’v I ‘ f.‘
Da1t141" 7OLD (1'5) 311
z“. , . / 1“ t ‘ J." .,
I 0U. O \ u S ) 1 l O

-1. '
1T3. L l O 53 BS

0 I
\Vel]_2‘S
,, . '-' 4. 4-
eLDerlmeDLaI L L

n. .1 Q l n . O ‘ N ‘3‘,
1'1- S'CO” ”2‘ {‘8 (35102310 bile

of recrxu Lment (If)
Div DinerimentDI

‘ ..' —A. 4‘.

Loan ShQGJh d soriDiDDt
twentv Clbss I (I's) an
Cless II (II‘s) suhfieci
coszetsiOD ratios es \
Dost- C’“QF'JOUTL_ tesis
less recxtuflxmnrt than J

H4
' |

wwwm 01'?
ruitme at (I.) scores
\VGIFF DeDQI‘ ‘
“Dell. t (Iow-

8101/1 o o o o o o

D
o u.

C)

?-<

The

VA.
DJWI'C“

) (‘1‘
1:3
c ’r
r-H .
('1‘
(.4.
D
'\
6+“

0 2
0:5

.. 1. .
sreecn D
scores
.1 subjects

Le

the s'

the sneech noise 6
1or'Um>Di:
e

”TV-€38. 3

78.11“.

H.
v
I"

0-]
\

record

C \ 73“ 1"» ‘fv {N C‘ (‘1 q [
' (,1. . I ‘_. \_.4 LV.‘ L .
u

(

\-
o
o
o
o
o
O
O
o

In“ DDG

..

r- .1-m

(J.I\J \FC}" ‘1’“:
(3 CO wwh-v’wmflw 071 (C)

l \ ,
\Q) 0 O o 0

e10

. k

0
for t'

a" -p "
sco es 1 tie
xteen

for tDe

I. I" ’1
as tDe Dle— “DD
7' . . 4—
lass 1 DDmr seDus
__..\.
S .1.) O I O O O O O

lk‘K
we subjects
3t (hifih-Dain)

) by ihe aDDlitude
I O C O C C O C C U

A; ' {7:

1’1
L,(]l

C) L)

'\\‘
L...J

L1-

ADDeDD1:

L1

L4
0

Id
?e
of
W O

. .1- ' -r‘
0111,34”

(~43 N ‘4 (1‘
QDOHQGL, mi

-L‘
LDC

rds a

”3

~ ~—
1

m11
'-
'L.

, r"
I - - I ‘ I A. ‘l'w'r'r‘ "‘I
0-- -5 1.14.; LL.)

0 ’t :1 G

‘3’
2' 3
C]
Q
,J
("F
a
(4‘
i...) u
9
»—‘
O
fi
5

.

._)
KW

Edwin—S '1'};

nc Inger:

U rum-r U

.ingvv

33"011‘7‘. S

‘r
‘5-

DT
.L
I
l
.L
0

Yr“

‘1;

o
".

1
.L— ‘

I} ~"'
, \
c..... “f

'._._l
O
Q‘\

The idea that amplitude compression can aid persons
with loudness recruitment in their perception of speech
is neither novel nor new. Huizing (1948, 1952) and
Huizing and Reyntjes (1952) implied that speech
discrimination was impaired by loudness recruitment because
of increased sensitivity for differences in intensity and
thereby normal loudness relationships among speech sounds
are altered. Caraway (1964) and Caraway and Carhart (1967)
conjectured that persons with recruitment, therefore, may
not need as large a dynamic range as do normal listeners
in order to take full advantage of speech signals. They
hypothesized that speech with reduced dynamic range (i.e.
amplitude compressed speech) would improve Speech
discrimination in subjects exhibiting loudness recruitment.
Their hypothesis was tested by comparing the discrimination
scores for compressed speech in four groups of subjects
whom they assumed to differ in their magnitude of loudness
recruitment. The groups demonstrated only slight
improvement for the compressed speech, and it was concluded
that the technique offered no important advantage over

linear amplification.

.1

'he present study extends Caraway and Carhart's
research by first operationally defining loudness recruit-
ment and then testing their original hypothesis within
this constraint.

A necessary requirement for conducting the
experiment was employment of a method for quantifying
loudness recruitment. An alternate binaural loudness
balancing technique, employing Speech noise signals, was
devised for this quantification.

Specification of the performance characteristics

of the amplitude compression system was also of primary

importance.

Purpose of the Study

This study sought basic information concerning the
relationship between loudness recruitment and the
discrimination of amplitude compressed speech. The purpose
was to compare Speech discrimination performance of two
groups that were heterogeneous with reSpect to degree of
loudness recruitment. Speech Signals were presented at
two levels of amplitude compression and also with no
compression. The following questions were formulated to
delimit the research:

1. Does amplitude compression improve Speech

discrimination in persons who exhibit loudness

recruitment?

2. IS the amount of improvement in Speech
discrimination related to the amount of the
loudness recruitment?

3. Is the amount of improvement in Speech
discrimination related to the degree of the

amplitude compression?

Importance of the Study

 

Design Specifications of hearing aids involve
numerous compromises necessitated by the interaction of
distortion produced by the aid and by the response
characteristics of the patholosical ear. Appr0priate
hearing aid design depends, or Should depend, on an
understanding of these interactions.

In reference to the relationship between the
characteristics of amplitude limiting systems and hypoacusic
speech discrimination, Lynn (1962) commented that the
"scanty" evidence is contradictory and indicates the need
for additional study. Furthermore, Caraway's research,
which was conducted in 1964, is the first and only
systematic, empirical investigation employing amplitude
compression (as distinguished from amplitude limiting).

The idea that amplitude compression can offer
znecruiting ears special enhancement in their discrimination
Of Speech signals has, therefore, received recognition but

Onlqv'very limited study. In this context, recruitment has

it

never been quantified and many basic questions remain
unanswered.

Information obtained from the study is viewed as
having possible application in increasing the hypoacusic's

potential success in utilizing amplification.

CHAT TER II

BACKGROUND II‘JT‘CRI-LATICI‘J AND LITELATLR; HEVIE “J

This chapter describes the pierformance Chara eter-
istics of ampli tude controllins devices and summarizes
their effects on Speech discrimination. Amplitude control

as a solution for loudness recruitment is als odiscussed.

‘erformance Characteristics of

L—l
v
1.,

 

Amplitude Cont rollin; Devices

 

(‘l "I

Three important ieatures o: amplitude controllins
devices are input- ~output relatiomi hips, time constants,
and problems associated with the control.

Input-Cutput Relationships. In linear amplifiers

 

input-to-output intensity ratios remain essentially
constant. For example, the output sound pressure will
increase 10 dB if the input signal is increased 10 d3.
moreover, this linear ra atio is maintained over most of the
Operational range of the instrument, xcept at very hifih
input levels. Input above this level is usually accompanied
by marked distortion of the waveform of the output signal.
Electrical engineers have sought linearity in
eamplifiers for many applications. There are, however,
expecific Situations which necessitate nonlinear amplifiers

or'amplitude controlling devices (Appendix A). These devices

KR

O\

O

restrict either the dynamic ranse or the peak output of
amplified Signals via nonlinear input—to-output pressure
ratios. They are used routinely in the recordina,
transmission, and reception of audio signals, and recently
they have been incorporated in hearing aids.

Two types of amplitude controlling devices ar
amplitude limiters and amplitude compressors (Langford-
mith, 1952). Limiting amplifiers are linear for low-
input signals; however, gain is reduced when the input
exceeds a set value. All inputs above this critical level
Show a fairly constant output. Compressors, operating on
a somewhat different principle, are amplifiers with input-
output ratios such that output pressure is inversely
preportional with input Sisnal strength. Input—output
functions for limiters, compressors, and linear amplifiers
are Shown in Figure 1.

In linear amplification (curve A-B—C), gain is
constant over the entire input range. For each increment
in the input Signal, an equal increment is observed in the
output signal. This relationship yields a linear slope at
an angle of 45 degrees.

On the other hand, amplitude controlling systems
<io not have constant input-output ratios. Curves A-B-F and
ti-B—G illustrate the output of two limiters. Both curves

are linear below a critical level (point B), but increments

Do-

80-

telative 70-

60.. B G

 

 

Eisrre l. fiain charactcristi,s c1 limiter and couDrc380"s.
(A-?-C) linear amplifier 1:1 ratio, (A-C) compressor 2:1
ratio, (A-E) compressor 3:1 ratio, (A-B-r) limiter 3:1 ratio,
(A-3~G) limiter 1:0 ratio (fro: Caraway and Carhart, lgc7,

D. 1&26).

in the input signal above this point produce relatively
smaller increments in the output signal. Curve A—D—G is
actually an idealized limiter with output gain reduced by
the amount that the input exceeds point B. Curve A-B-F
Shows the limiter function usually found in clinical
application. These functions have fixed maximum output
which tends to eliminate the distortion that accompanies
linear amplification at very high input levels.

Curves A-D and A-E are compressor functions with
two-to-one and three—to-one d3 of input-output gain
respectively. In amplitude compression, the shape of the
waveform remains relatively intact while amplitude
relationships are modified according to the input-output
ratio.

The net effect of amplitude compression is
dynamic range reduction. The amount of the reduction
is determined by the input-output ratio of the amplifier.
Reduction produced by one—to-one, two-to-one, and three-
to-one compression ratios for Speech signals having a
30 dB dynamic range are shown in Figure 2. The peak
components of the Speech signals, in this example, are
equated at 30 dB sensation level. Dynamic range reduction
under the two-to-one ratio is 15 dB, under three-to-one
the reduction is 20 dB. This means that the peak

components are held constant for each ratio, whereas the

J _
Ranzc of 20 ' '
Sensation 10
Level 0

 

1:1 2:1 3:1

Input—Output Ratios
Figure 2. Dynamic range reduction produced by one-to—one,
two-to-one, and three-to-one amplitude compression for
speech with a 30 dB dynamic range.
fainter components are amplified relatively more. The
term "compression", therefore, refers to dynamic ranee
reduction and should not be confused or equated with
presentation level per se.

Time Constants. Amplitude control can be thought
of as auto-regulation of gain based on input signal
strensth. Generally, the control action results from
negative voltafie feedba h. For example, increments in
input signal strength cause relatively more negative
voltase to be applied to the amplifier, resulting in less
gain. Decrements in the input signal strcnjth result in
less nemative feedback, producing relatively more
amplification.

The time needed to accomplisi gain increments or

n

decrements are called ”time constants". The lapse beiore

gain reduction is called "attacm time", and the lapse

before gain is restored is called either "release" or

g...
:4
(D
C’)
(D

"recovery time". concepts are applicable both to

lO

circuits which use control voltaee to vary the gain of a
linear amplifier and to circuits that Operate on nonlinear
components, as does the Weiss unit.

There is no standard for specifying the time
constants of amplitude controlling devices. The percentage
change in gain upon which these intervals are based are
usually not specified in literature reports. The
specifications utilized in this study, as well as the
previous study by Caraway (196M), were suegested in 1949
by Grimwood. He preposed computing the constants upon
ninety percent completion of gain change because the value
provided a realistic estimate of the function and because
it was manageable from the standpoint of accuracy of
measurement.

The only systematic investigation of the influence
of time constants on speech discrimination was conducted
by Lynn (1962). Results, published by Lynn and Carhart in
1963, indicate that the time constants influence both the

audibility and intelligibility of Speech.

Problems Associated With

 

Amplitude Control
ApprOpriate selection of the time constants can
minimize the undesirable products of amplitude control.
This section is devoted to a discussion of these problems

and their amelioration.

ll

Thump. Audible pulses which result from fast
operating time are called "thump" (Naxwell,l9b7; Grimwood,
1949; Singer, 1950; and Ancona, 1956). This is a d.c.
pulse which results from the sudden change in gain.
Fortunately, the pulse is usually low frequency and can be
eliminated by low frequency filtering (Marcus and harcus,
1965).

Program Gaps and Pumping. Program gaps are caused

 

by Slow release times on occasions when a faint sound
follows a loud sound. The loud Signal triggers an

amplitude control response which can obscure a weak

signal when it is in effect. Until the amplifier "recovers",
its performance is analagous to the refractory phase in
neural stimulation. "Gaps" in the program can be eliminated
by using faster release times; however, "pumping" may

result when this time is too fast. Pumping is excessive
attack and release and occurs as a result of the rapid
fluctuations in running speech (Maxwell, 1947; Grimwood,
1949; and Poliakoff, 1950).

Excessive Noise level. Excessive noise is created

 

by ambient noise being amplified at higher than normal
relative levels. This noise is analogous to the fainter
speech components in that it is amplified proportionately
more than the stronger components. Several writers have

commented about this undesirable concomitant feature of

12

amplitude control (Davis et. al.. 1947; Parker, 1953;

Poliakoff, 1950; Rutherford, 1957; and Caraway, 1964).

Effects of Amplitude Control Cn

 

Speech Discrimination

The effects of three methods of amplitude control,
namely, peak clipping, amplitude limitation, and amplitude
compression, will be considered in this section.

In peak clipping the amplitude of the sound wave is
eliminated above a certain value. The degree of clipping
can be specified by the fraction of maximum amplitude that
is passed. Several writers have reported on methods of
peak clipping and the effects produced by this method of
control (Licklider and Pollack, 1948; Tolhurst, 1959;
Velichkin, 1962; Gustavsson, 1963; and Pyron and Williamson,
1964).

Normal hearing subjects are able to understand
Speech which is dramatically clipped. Evidence of this
ability is provided by Licklider and Pollack's statement
that

the so-called dynamic characteristics of speech are
not of vital importance for intelligibility. It is
apparently just as well to reproduce all the
fundamental Speech sounds (or what is left of them
after infinite clipping) at the same intensity as it

is to preserve their natural intensities (1948, p. 49).

This implies that intensity variations are probably not

the basic cues in Speech discrimination. Their statement,

13

however, Should not be interpreted as implying that
intensity distributions of normal Speech have no role in
recognition.

Davis et.al. (1947), as a part of the Army and
Navy aural rehabilitation programs, reported on a master
hearing aid with both peak clipping and limiter action.
The superiority of limiter action over peak clipping was
shown in a series of speech discrimination tests using
Harvard PB-SO stimulus material in three normal—hearing
subjects. Similar results were observed in six hard-of-
hearing subjects.

Hudgins et. a1. (1948) built a wearable hearing
aid with limiter action. This instrument was compared,
via articulation tests with six hard-of-hearing subjects,
with the master hearing aid just discussed and with two
commercial hearing aids. The subjects were trained
listeners who had utilized amplification for at least one
year. Comparisons were based upon percent correct
discrimination of Harvard PB-5O word lists (presented via
monitored live voice) for successive input levels.

Results were summarized by the following statement:

These results indicate clearly both the feasibility

and the desirability of limiting the output of hearing
aids by means of compression amplification. Furthermore,
it is demonstrated that limiting the power output by
means of compression amplification not only protects

the ear but at the same time reduces distortion to a
minimum, thus maintaining in most cases the maximum

level of performance over a wide range of speech-
input levels (1948, p. 253).

14

Several researchers have reported dramatic
improvements in speech discrimination for hearing aids
with amplitude limitation when compared with conventional
hearing aids (Fournier, 1951; Pestalozza, 1953; Portmann
and Portmann, 1961; Rice and Flemming, 1968, Bizaguet,
1968; and Flemming and Rice, 1969). Fournier (1951) and
the Portmanns (1961) Showed articulation functions for
patients who demonstrated superior understanding for
limited Speech as compared with linear amplification.
Pestalozza (1953) compared hearing aids with peak clipping,
with volume limiting and with linear response. Superior
scores were obtained under the volume limiting.

Edgardh (1951) observed that dynamic equalization
was not obtained between the vowels and consonant sounds
unless amplitude control functioned over the entire
Operative range. Working with extreme limitation, he
found that speech was as intelligible as untreated speech,
although the subjects observed quality differences.

The effects of peak clipping and amplitude
limitation in the presence of white noise were studied
by Kretsinger and Young (1960). These investigators
compared the effects of two degrees of control, achieved by
either method, on the discrimination of thirty normal-

hearing subjects. The subjects exhibited superior scores

 

'_l

U)

15

for the limited PB-5O word lists (C.I.D. Auditory Test W-ZZ)
than for either no control or peak clipping.

It has been asserted that the amount of amplitude
control preferred and needed by persons with different
types of hearing loss varies considerably (Silverman and
Harrison, 1951; Poliakoff, 1950; and Rice and Flemming,
1968). Parker's (1953) findings support this assertion.
He investigated the effect of amplitude limitation in
ten persons with various amounts of sensorineural hearing
loss by constructing articulation functions for Harvard PB
words at sensation levels of 6, 16, 26, and 36 dB. Results
indicated considerable individual subject variability in
the benefit afforded by limitation. Of the ten subjects,
nine did obtain higher intelligibility with amplitude
limitation at one of the sensation levels. Three had
superior discrimination at all levels; six had superior
discrimination at some levels and poorer scores at other
levels; only one failed to benefit at all test levels.
Notably, improvements ranged from 2 percent to 50 percent
and the maximum help seemed to be at the lower sensation
levels.

Lynn (1962) systematically investigated time
constants of limited speech in groups of subjects with
otosclerosis, labyrinthine hydrOps, and presbycusis. The

limiting was achieved with an instrument built from hearing

l6

aid components. Time constants were found to influence
Speech perception of Harvard PB word lists. Labyrinthine
hydrOps subjects performed Optimally with shorter release
times, and otosclerotics performed as well when no
limitation was used. Lynn believed that time constants
could be selected which yield Optimum intelligibility and
that these values do not appear to remain constant for
all types of hearing loss.

It is important to note that the reports involving
limiting demonstrate advantages for this type of amplitude
control when compared to abrupt peak clipping and no
advantages over speech without amplitude control.

As previously mentioned, the only systematic
investigation of hypoacusic speech perception using
amplitude compression per se was conducted by Caraway
(1964). She investigated the effect of two degrees of
compression compared to no compression using twelve
subjects in each of the following catagories: labyrinthine
hydrOps, labyrinthine otosclerosis, presbycusis. and normal
hearing.

Amplitude compression was achieved with the same
compressor that was used in the present study. The
instrument had compression ratios of two-to-one and three-
to-one and it also functioned as a linear amplifier, called

one-to-one compression amplification. Performance

17

Specifications of the instrument include a flat frequency
reSponse from 250 to 5000 Hz and time constants for all
amplification conditions of 1.5 msec rise and l msec decay.
Caraway determined speech reception thresholds
for Spondees under each compression condition. She also
constructed articulation functions for CNC words
(Northwestern University Auditory Test No. 4) under each
condition at sensation levels of 0, 8, l6, and 20 dB.
Performance in speech discrimination improved only
minimally for the compressed speech in all her subjects.
Also, no important differences were found among the three
degrees of compression when comparisons were made at a

given sensation level.

Amplitude Control as a Solution

for Loudness Recruitment

 

Persons with recruitment exhibit a reduction in
the range of comfortable loudness for Speech signals.
This means that the decibel difference between the
faintest audible sound and the strongest tolerable sound,
i.e. the dynamic range of hearing, is reduced. At the same
time, intensity variations encountered in the daily
acoustic environment are great enough to tax even the
normal ear's wide range of comfortable loudness. Noderate
reduction in the comfort range may lead to problems in

hearing faint sounds or annoyance from loud sounds.

.m.
l --‘

 

 

 

18

Conventional linear amplification may not be ideal
for persons with recruitment or a tolerance problem because
it must be Operated at high gain in order to make the faint
sounds audible. When these sounds are amplified to an
audible level, the loud sounds may become intolerable.

With this mode of transmission, the only solution is gain
reduction, thus sacrificing intelligibility of the fainter
sounds.

Several people have expressed a special need in
hearing aids for peOple with recruitment or for similar
problems associated with tolerance (Groen, 1951; Neuberger,
1954: and Portmann and Portmann, 1961). Davis and
Silverman (1970) contended that persons with loudness
recruitment encounter difficulty in linear amplification
because of the abrupt transition from hearing little or
nothing to hearing very loud sounds. In other words,
recruitment both disrupts normal loudness relationships
and sharpens differential sensitivity (Huizing and
Reyntjes, 1952; and Caraway and Carhart, 1967).

Pestalozza (1953) compared Speech reception
thresholds and Speech discrimination measures for peak
clipping and amplitude limitation. He found that
limiting was the method of choice for controlling the

maximum power Of hearing aids and that linear amplification

19

increased distortion and reduced discrimination when marked
recruitment was present.

A specific type of articulation function,
"regression", which appears in sensorineural hearing loss
when recruitment is present, has been reported by several
writers (Dix, Hallpike and Hood, 1998; Eby and Williams,
1951: Huizing and Reyntjes, 1952; and Schultz and Streepy,
1967). This function has increments in discrimination
for increments in sensation level up to a point, but the
discrimination decreases fairly rapidly for higher
sensation levels. It has been suggested that Optimum
amplification would be provided by output limitation
corresponding to a person's Optimum range Of intelligibility
(Shutts, 1950; and Huizing and Reyntjes, 1952).

Design characteristics and an excellent rationale
for using compression in amplification for recruiting ears
were reported by Aspinall (1951) and by Rice and Flemming
(1968). However, experimental results are not presented
in either report. Other writers have specifically
suggested compression for persons with a narrow range
between the thresholds of intelligibility and tolerance
limits (Ashton, 1951; Poliakoff, 1950; and Nenzel, 1966).
Again, however, eXperimental evidence is scanty and the
effects of recruitment on the perception of amplitude

controlled speech is not known.

§m§£1c

Amplitude controlling devices, limiters and
compressors, can be specified by their input-to-output
dB ratios. Their performance as well as their effects on
signals they transmit is markedly different. Amplitude
compression acts to reduce the dynamic range Of Speech
signals, whereas amplitude limitation tends to limit
the maximum levels Of the signals.

Although the time constants Of limiters were found
to influence audibility and intelligibility, there has
been no systematic study Of these constants in compressors.
There appear to be concomitant problems associated with
amplitude control. Fortunately, most of these difficulties
can be minimized or eliminated by apprOpriate selection of
the time constants.

Only one study (Caraway and Carhart, 1967) which
investigated the effect on speech discrimination of using
amplitude compression is reported in the literature.
Several studies using amplitude limitation have been
reported. These indicate that limitation appears to be
preferable as a means Of limiting the maximum output of
hearing aids when compared with abrupt peak clipping.
Extreme limitation does not seem to effect adversely speech
understanding in normal hearing persons. However, when

clipped and limited speech are immersed in noise, normal

21

hearing subjects understand the limited Speech better.
Other studies have shown that sensorineural subjects vary
greatly in the amount of benefit they achieve through
limitation and that time constants influence discrimination.

The only study involving compression failed to
demonstrate any important advantage for the amplitude
controlled Speech over Speech with no amplitude control.
It is important to note that the reports involving
limitation demonstrate advantages for the control when
compared to abrupt peak clipping and no advantages over
speech without control.

Finally, the question remains regarding the
differential effect on Speech discrimination as a function

of the degree Of loudness recruitment.

CHAPTER III

EXPERIRENTAL PROCEDURES

This study investigated the ability Of persons
with different amounts of loudness recruitment to discrim-
inate amplitude compressed speech. An alternate binaural
loudness balance technique provided a criterion for
placing subjects into either a partial or a complete
recruitment group. Subjects were then tested under two
degrees Of compression and a condition of no compression.
Two discrimination scores for CNC monosyllabic words were
Obtained for each subject under the three compression
conditions.

Amplitude compression was accomplished with a
prototype instrument built by Er. Erwin Weiss and supplied
by the Beltone Electronics Company. This compressor had
input-to-Output amplification ratios of two-to-one and
three-to-one. It also functioned as a linear amplifier;
this condition was called one-tO-One amplification.

The test battery contained both preliminary
audiometric evaluation and Speech discrimination measurement.
This included pure tone air and bone conduction thresholds,
Speech reception thresholds, Speech noise alternate

binaural loudness-balance test (ABLB), and Speech

22

23

discrimination tests under each of the three compression
ratios.

For Speech discrimination testing three permutations
of Northwestern University Auditory Test No. 6 (lists 11
and III) served as stimulus material. The two scores at a
given compression ratio were based on one arrangement Of
List 11 and one of List III. Each list was recorded at
the three compression ratios in such a way that the average
peak levels of the output signal were equivalent. That
is, the Speech Signals for each condition were equated at
their peak levels, and long term dynamic ranges below
these levels differed from one ratio to another. For
example, the uncompressed dynamic range of the speech
stimulus material was approximately 24 dB. Under two-to—
one compression this range was reduced to 12 dB and under
three-to-one it was reduced to 8 dB.

A pre-test CNC list was presented at the outset
and again as a post-test to investigate practice effects

and time dependent influences.

Sppjects
Thirty-six subjects, constituting two groups Of
eighteen each, were selected for the study. Their age

and sex distributions are shown in Table I.

01’,

TABLE I

Age and sex distributions for the two groups

 

 

 

Recruitment Age in Years Sex
Distribution
Group Mean Eedian Range Male Female
Partial 49.0 50.1 31.2-68.2 5 13

Complete 52.7 54.7 21.5-68.3 ll 7

 

25

Preliminary selection of subjects was based on the
results of audiograms available at the offices of
Dr. Sanford C. Snyderman, Dr. E. H. Bersendahl, Dr. Charles
S. Giffin, Dr. John Thomas, and hr. Patrick Carter in
Fort Wayne, Indiana.

Each met the following criteria for inclusion into

the study:

1. All subjects were medically diagnosed as
having unilateral, senso_ineural hearing loss.

2. Better ear within normal limits (air and bone
conduction thresholds no greater than 25 dB re
ISO-1964 Norm) for 500, 1000, and 2000 HZ.

3. Intraaural Speech reception threshold
differential of 25 dB or greater.

0. No air-bone gap or no response by bone
conduction at the maximum limits of the
equipment.

5. Sufficient residual hearing in the Speech
frequencies of the test ear to allow speech
discrimination assessment of 24 dB sensation
level.

6. Eighth grade or higher education.

7. Hearing loss acquired after the develOpment

of English language.

26

The restrictions of a normal better ear and a 25 dB
intraaural Speech threshold differential were necessary for
administering and obtaining meaningful results with the
Speech noise ABLB test. The eighth grade minimum education
level increased homogeneity with respect to test vocabulary.
Possible contamination due to faulty language and due to
foreignism was controlled by requiring all subjects to
have acquired hearing loss after the develOpment of Enalish
languase.

Median pure tone thresholds of the two groups are
shown in Figure 3. For clarity, bone conduction responses
are omitted. The air and bone conduction results for both
groups were interweaving, excepting some subjects whose
bone conduction reSponses were beyond maximum limits of
the audiometer.

AS anticipated, the groups exhibited similar
audiometric configurations bilaterally. Median nontest,
better ear results are remarkably alike. Despite differences
in the low frequency region, both groups Show unilateral
hearing loss with gradually downward sloping configuration
in the test ear.

A comprehensive summary of the subjects reSponses
to preliminary tests is presented in Appendix B. Pure tone
thresholds, pure tone averace thresholds, and Speech

reception thresholds are shown.

. .H‘GO

some one sow crepe cam mcaoamosrw mo mousse man .sxogm chm mpaogmosmp sofleo5ccoo
than ems >Hso .hpflnsflo son .mezopa psosefl5poms neon mom «ocfla pogmcpv ado mosses
ore was Aosfla pflfiomv mam snow opp new mpaommosre oCOPIQSSQ swamp; .m marten

Amuv mesonsopm

000$ 000$ QOON oooa o
— ,Ll _

'- 0
UN

0mm coca coo: ooom coca com 0mm
_ . c . .

4. o:

l -
‘1

u flog

f 00 B

 

 

 

 

 

 

 

 

 

 

 

 

TV.
l .. Io, u
Ci I4 J] I..' III C J
11 I Ion “Um
. L
J '0 00 AU
j Q S
T.. U.
\1 (x.

I low mum
r , p
I. l"/ I5. [om AHA
r ’ (a
ll.‘ I. I, I

' )
"’II 1. I.” filom T.
’ ’I I" U

I- I “| 'I “‘l' I'
I lu‘ I. In ’Iu‘

I.. I OH D
5..“

l 'O

ssosw ogmacroo ssonc asseQSH

 

 

Recruitment Duantification

 

A necessary requirement for conducting the
experiment was the develOpment of a method for quantifying
loudness recruitment. Although Jerger and Harford (l960)
and Jerger (1962) contended that the pure tone ABLB test
is the most valid existing measure of loudness recruitment,
results of this test are somewhat unwieldy. Attempts to
relate recruitment to other variables have suffered because
of the difficulty in quantifying recruitment by this pure
tone procedure, e.s. Clemis and Carver (1967). An
alternative procedure using Speech stimuli may offer some
advantages.

Harris et. al. (1952) stated that "The practical
Significance of recruitment is largely in Speech reception"

(0. 108). In an investigation of recruitment for speech,

Q

thi roup of investigators found their subjects could

U)

\
’

(T

readily make loudness matches for Speech in two ears of
unequal ability and that the matches were as easy and
precise as those for pure tones. It was also found that
"recruitment for Speech followed a course parallel to
that for pure tones and roughly intermediate among the
curves of recruitment for pure tones in the Speech ranse"
(1952. p. 132).

Studying the effects of noise eXposure on

loudness-balance and intelligibility, O'Neill (1954) used

29

Speech material for pre- and post-exposure loudness-
balances. Mean values of the matches indicated that his
normal hearing subjects were able to make accurate
judgments with connected discourse serving as the stimulus.
O'Neill concluded that,"Apparently speech (connected
discourse) can serve as adequate stimulus material in
loudness-balance matches" and that "there was little
intra- or inter-individual variability in such judgments"
(1954. p- 6).

Specific rationale for employing speech noise for
the ABLB test stimulus include the following: (1) literature
reports that subjects can adequately balance Speech—type
stimuli; (2) the close relationship between Speech noise
and the dependent variable: and (3) because the Speech—
type Signals are capable of demonstrating growth of the
loudness function over a relatively broad auditory area.

Speech Noise ABIB Test. Stimulus parameters and
test procedures are Similar to those suggested for the
pure tone ABLB by Jerger (1962). Identical Signals are
presented to both ears alternately. The Signal was fixed
at levels of 20, 40, and 60 dB above the good ear'S Speech
reception threshold and the subject adjusted the intensity
at the poor ear for each presentation level until the

signals were judged equally loud in both ears.

30

The procedure for administering the test described
here calls for fixed sensation levels in the good ear and
variable intensity in the poor ear until the balance is
achieved for each reSpective presentation level. Jerqer
and Harford (1960) and Jerger (1962) recommended fixing
the signal at sensation levels of 20 and 40 dB above the
poor ear's threshold and varying the intensity at the good
ear; however, they reported that either system yields the
same results. Their rationale for these recommendations
include standardization and clinical utility. With
respect to the presentation levels, they suggested that
"the 20 dB level sets at the question of recruitment near
threshold and the #0 dB level is usually high enouqh to
tell you whether recruitment is complete" and that "it is
very seldom that you can go much hiaher than 40 dB above
threshold on the bad ear anyway" (1962, p. 142).

The decision to fix the three sensation levels at
the good ear was governed by the desire to obtain information
about the loudness function over a broad span of sensation
levels. Because the magnitude of loudness recruitment is
relative and varies as a function of stimulus presentation
level, a procedure based on the good ear's threshold level

Inay lead to somewhat different absolute results than those

obtained by classical procedures.

31

Specific siinal parameters and procedures included
the following:

1. Signals alternated automatically.

2. Signal duration was 500 msec.

3. Signal rise and decay time was 50 msec.

M. The intensity was always fixed at the good ear
and varied at the poor ear.

Each subject controlled the intensity at his
poor ear by a hand-held switch.

6. The intensity was fixed at sensation levels of

20, 40, and 60 dB at the good ear.

An Allison 22B clinical audiometer was used to
administer the test. In order to obtain a speech noise
signal and the capability of alternating Signals to a
subject's two ears, the test signals were recorded on
magnetic tape and played-back on the audiometer's "stereo"
magnetic tape recorder.

The array of equipment shown in Figure 9 was

utilized for recording the test stimuli.

 

 

 

 

 

 

 

 

Grason-Stadler Grason-Stadler Ampex AG
901-3 ]______ 929-E 350-2
Noise ' Electronic Tape

Generator Switch _‘ Recorder

 

 

 

 

 

Figure Q. Schematic diasram of the Speech-Noise Alternate
Einaural Loudness Balance recording equipment.

The noise generator (Grason-Stadler 901—3) produced
Speech noise which was transmitted to the electronic
switch. The switch was adjusted at a 50% duty cycle,

50 msec rise and decay time, and a 1000 msec period.
These adjustments resulted in brief bursts of the speech
noise being transmitted (switched) alternately to the two
separate channels of the Ampex Tape Recorder (AG 350-2).
When the two channels of this recording procedure are
transmitted separately to the left and right ears, the
Signals are perceived as alternatinfi. Signal parameters
were monitored with a Tektronics (564 B) Storage
OscillOSCOpe.

Two minute continuous segments of Speech noise were
recorded at the beginning of each channel for calibration.
Spectral analyses of these Signals are Shown in Figure 5.
The curves Show that the frequency Spectrum of the recorded
signals conforms to the long-term average Spectrum of
Speech Signals (Denes and Pinson, 1966; also see Appendix A).

Int:rpretation of Speech Noise ABlB Test. Three

(3

1')

possible outcomes of pure-tone AYES tests are "no
recruitment”, "partial recruitment", and "complete
recruitment". Figure 6 illustrates these hypothetical
outcomes. The test Signal is initially fixed at a level
of 20 dB above the good ear's threshold (01) and the

alternating signal is turned on. Because the good ear is

.nasshflw snow

0
’3
j)

J
O
O
O
\ r 3

000:
.

mama

pawn

0;
or—{
m
\J
r.
H
H
O
(Ll
Q)
.r‘;
r.
Cm
C)
(’3
4..)

4 HoCCQKQ

 

air

lllalnnunllllll

aosoSQosh .m wedge

oom 9mm

lelllllllx

OIIIIIIIIO

ouuuu“u\.e

x

ION...
IWHI
fies-

Um...-

 

U.

P. .7

10‘»:
LL

Spelt?

C"?
-uJU

3A

.Asde goo: 4 “ado root .Qv

kV
new oocnans moosugoa HSQSCCHQ owdflhowam one we mpafimoa heapescogssfl
podnet Hmcoawflprsp one wsflwdswmzaafl mefltacxe amow+eswoi>x .w cyztm;

PcheHSSopm PQmEPHSMoox PeoseHSSop:
meoaesOo Hmflppwm oz

M (ow
_ l/I/l/IV 0%

.00

- d \ Mp Cw

W Jr]! (fl .OU HG>QH

 

 

 

 

 

 

 

 

 

 

Q As OaN dwtfixdmfiw
/ / .or
. (Q Aw C om

 

OH

 

 

 

 

 

I;

35

normal in the illustration, sensation level and hearing
level are the same; however, it Should be noted that
sensation level is in reference to the individual's hearing
status (See Appendix A Re: Hearing Level and Sensation
Level). The subject adjusts the Sienal at his poor ear
until he judges the signals were equally loud (X1). Next,
the subject repeats the balancing in his poor ear for
signals presented at MO dB and then at 60 dB.

If the subject perceives Signals presented at
equivalent sensation levels in each ear as equally loud,
he is not exhibiting recruitment. This means that an
increment of say 20 dB in the good ear is accompanied by a
20 dB increment in the poor ear to effect a loudness
balance. On the other hand, partial or complete recruitment
occurs when a smaller increment is needed for the balance.
In fact, if.at the 60 dB sensation level, the subject
adjusts the intensity at the poor ear eoual to the intensity
at the sood ear (in equal hearing levels), his performance
is labeled as "complete recruitment". Balances which
occur between the "complete" and "no” recruitment responses
are labeled "partial" recruitment. Jerger (1962) reports
€1.10 dB "margin of error" for interpreting pure tone ABLB
dat61. Presumably this error is related to dispersion of
the loudness balances. Preliminary results in this

eXDEBriment, with Speech signals, indicated that subjects

36

were able to balance the sianals within a much smaller range.
Accordingly, the margin for interpreting the balance data

of this study was reduced to 5 dB.

The two groups of subjects used in this study
conformed to the above specifications for partial and
complete recruitment. That is, when the mean of balances
was rounded to the nearest 5 dB, eighteen subjects

exhibited partial recruitment and eighteen exhibited

complete recruitment for the Speech noise Signals (loudness

recruitment is defined in Appendix A).

Speech materials

Three Speech tests were required by the eXperiment:
(1) determination of sensation levels for presentation of
the compressed Speech; (2) measurements of discrimination
performance under compressed Speech; and (3) pre-test and
post-test to assess base level performance and time
dependent influences.

Speech reception thresholds, based on Spondee
words, were used in determining the presentation levels
.for the compressed Speech. Magnetic tape recordings of
the CID Auditory Test W-l (Hirsh et. al., 1952) were
recuarded by a male speaker (Dr. William F. Rintelmann)
Witli General American dialect. The speaker monitored the

leveal of the two syllable peams of each word at 0 dB VU

metexr deflection (:’2 dB). List A, one of four lists

37

used in routine clinical testing at Michigan State
University, was arbitrarily chosen as the stimulus material
for the preliminary speech test.

Northwestern University Auditory Test No. 6
(Tillman and Carhart, 1966) was used to assess Speech
discrimination under the three compression ratios. This
test has four lists of 50 monosyllabic, CNC words derived
from lists deveIOped and revised by Lehiste and Peterson
(1959 and 1962). The lists used in this study were
recorded on magnetic tape by the same male Speaker who
recorded the spondee material. Each word is preceded
by the carrier phrase, "You will say . . .". In recording
the words, the last word of the carrier phrase was
monitored at 0 dB VU meter level and the test item was
then said naturally.

Because each subject was tested twice at each of
the three compression ratios, a minimum of six different
test lists was needed. Previous investigations with the
N. U. Test No. 6 revealed that lists II and III yield
equivalent scores (Rintelmann and Jetty, 1968). Therefore,
these two lists were selected. Four additional lists
needed for the experimental conditions were constructed
1Wipermutine the two original lists; two permutations of
list: II A provided lists II B and II C and two permutations

0f Heist III A gave list III B and list III C.

38

Speech discrimination was measured at the beginning
and again at the end of the experiment by a pre- and post-
test. The pre-test acquainted the subjects with test
conditions and provided a base level for assessing initial
group differences. A comparison of pre- and post-test
scores indicated the magnitude of practice and other time
dependent effects. List I A of N. U. No. 6 was arbitrarily
selected for these measures. Of the fifty CNC words
that constitute this test, the first ten were recorded at
one-to-one compression, the next fifteen words at two-to-

one, and the remaining twenty-five words at three-to-one.

The Weiss Amplitude Compressor

 

Compression of the speech test material was
achieved with an instrument deveIOped by Mr. Erwin Weiss
of the Beltone Electronics Corporation of Chicago, Illinois.
Operating characteristics and calibrating information were
supplied by Mr. Richard Brander (1970). The compressor
functions to:
1. Divide input signals into three bands: #00 Hz
to 1000 Hz, 1000 Hz to ZOOOHZ, and 2000 Hz to
4000 Hz.
2. Provide symmetrical, nonlinear amplitude

compression in each channel.

\J)
\O

3. Provide independent control of the desree of
compression in each channel.
A schematic diagram of the compressor is shown in
Figure 7.

Input-Output Calibration. The instrument was

 

calibrated to the three compression ratios by adjusting
the Specified amount of compression for each channel while
maintaining equivalent gain in each channel. The capability
of adjusting both the amount of compression and the amount
of gain made available a wide array of conditions that
could be achieved with the instrument. The controls of
adjacent channels are not entirely independent, however.
This is particularly true when there is high gain combined
with a compression level and the filters do not cut off
abruptly. In this situation attenuated signals are
allowed to enter the wrong channel. Specific procedures
used in adjusting the controls of the compressor to the
three levels were as follows:
1. A Hewlett-Packard oscillator (QZOH-A) applied
a l V (RKS), 1500 Hz sine wave to the input
terminals of the compressor. A Beckman (6148)
Electronic Timer assured correctness of the
frequency output (E percentage error of 0.05%)

before and after the experiment. The internal

.mommmmeEoo weapflaesd meme; opp mo Edmumflp oamemzom .m mmsmflm

 

“cease;

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mmmeao>
.m Comm m phase 0 .w. gamma
Ase oooeuooomv m Hmccaeo ‘
o , r emaeaaaee
h. hmfiwflamfiw gm Coammmhano & Summ— . no.5
Use “em ooomloooav m Hmssmgo_ esp
mcfixfla - , ecfieopflzm
- , PSQCH
«a soammosasoo .fi seem

 

 

ANm oooaxooev H Hmcsmso

 

#1

volume control of the compressor was adjusted
to produce 1 V (RES).

With all compression controls at zero, and with
the 1500 Hz 1 V (HHS) input, the mid—channel
(1000 Hz to 2000 Hz) was adjusted to produce

1 V (HMS) at the output terminals. With 700 Hz
input (1 V HHS) the low-channel (400 Hz to

1000 Hz) gain was adjusted for output voltage
of 1 V (HHS). Similarly, with 3000 Hz input

1 V (RES) the high channel (2000 Hz to 4000 Hz)
gain was adjusted for 1 V (RES) output.

With a 1500 Hz input signal, the mid-channel
compression was adjusted until a 10 dB change
in output level was produced by changing the
input from 0 dB Re: 1 V (V) to -30 dB V. This
was for the three-to-one compression ratio.

For the two~to~one ratio the compression control
was adjusted to give 15 dB output change for

an input change from 0 dB V to -30 dB V.
Obviously, during the one-to-one condition a

30 dB reduction in output resulted from an
input change from 0 dB V to —30 dB V.

For each respective condition (i.e., one-to-one,

two-to—one, and three-to-one) step three above

£92

was repeated with a 3000 Hz input signal with
apprOpriate high-channel compression adjustment.

6. With 0 dB input, low channel (using a 700 Hz

signal) and high channel (using a 3000 Hz
signal) gain were adjusted until they were
equivalent in overall gain with the mid-channel
(i.e., the 1500 Hz signal) gain.

7. Steps number three to six above were repeated
for each respective compression condition (i.e.,
one-to-one, two-to-one, and three-to-one) until
the three were equivalent in gain and
compression.

Input-output relationships and the frequency
response of the comp-essor were monitored by the above
calibration procedures immediately before and after
compressing the stimulus word lists. Additionally a
calibration tape, consisting of two sections of recorded
pure tones, was constructed to determine performance of
the compressor during the actual processing of the words.

Section one of the calibration tape was used to
determine input-output relationships. It consisted of
three tones near the center of the bands passed by the
compressor channels (i.e., 700 Hz for the 400 to 1000 Hz
channel, 1500 Hz for the 1000 Hz to 2000 Hz channel, and

3000 Hz for the 2000 Hz to 4000 Hz channel). In constructing

43

this part of the tape, a 1 V (RES), 700 Hz tone which
produced 0 dB VU meter deflection was recorded for fifteen
seconds. Next, the level of the input signal was reduced
10 dB. A signal of fifteen seconds was again recorded.
In this manner, the input level of the signal was reduced
in decrements of 10 dB until a 30 dB Span was covered.
Similarly, 1500 Hz and 3000 Hz tones were recorded at
1 dB V and at decrements of 10 dB (i.e., -10, -20, and
~30 dB V).

Eleven pure tones (i.e., 200 Hz, 250 Hz, 500 Hz,
700 Hz, 1000 Hz, 2000 Hz, 3000 Hz, 4000 Hz, 6000 Hz, and
8000 Hz) constituted the second section of this tape.
These tones were recorded at constant voltage 1 V (RES) and
were used to determine the frequency response of the
compressor when the speech stimuli were actually compressed.

Rise-Decay Transients. Amplitude controlling

 

devices usually need intervals of time to achieve and
escape their control. These intervals are called attack
and release times or constants. Typically, these constants
are applied to circuits which use a control voltage to

vary the gain of a linear amplifier. Since the Weiss
Compressor Operated on non-linear components, its reaction
times are best referred to as rise-decay transients

(Caraway, 196Q).

M4

Transients of the compressor were determined by
observing its response to instantaneously initiated
sinusoids. A tone burst generator (General Radio 1319)
'nitiated the signals applied to the compressor and
externally trisfiered an oscillosc0pe (Tehtronics, 561) which
allowed observation of the compressor's performance.
Specifications reported by Caraway (1964) were adopted:

The rise time was defined as the time required for a

90 percentaae completion of pain chanae from the
instantaneous application of a signal to the compressed
equilibrium value. The de ay time was the time
required for the same percentaae of gain Change in

the return of the compressed gain to its former value
upon the instantaneous cessation of the input signal.

The compressor had a rise time of 5.0 msec and a
decay time of 2.5 msec. These values did not vary as a

function of frequency or of the compression ratios.

Harmonic Distortion. Energy measured in the first

 

two overtones was utilized to estimate harmonic distortion
(Caraway, 1964). For this measur ment, 1 V sinusoids with
calibrated frequency were applied to the compressor at
each ratio. The compressor was connected to a Bruel and
Kjaer 2107 Frequency Analyzer which enabled measurement of
the fundamental (h1)' the second harmonic (h2) and the
third harmonic (h3). Percent distortion was calculated at

each frequency by the formula:

Percent. SJ/ 3 + h2
Distortion

 

[1.5

Table II shows the distortion percentages.

Table 2. Percentage distortion of the second and third
harmonics of the Weiss Compressor for one-to-one, two-to-
one, and three-to—one compression.

 

 

Frequency

 

(Hz) one-to-one two-to-one three—to-one
250 3.31 3.06 5.63
500 2.78 6.39 1.50
1000 2.78 9.79 6.03
2000 1.95 4.77 7.96
3000 3.32 2.50 2.70
0000 2.06 1.81 1.61

 

4.72 4.2:

N
Q
Q

Mean Distortion

 

The distortion, although somewhat greater for the
two-to-one and the three-to-one compression ratios, was not
excessive.

Noise Level. In order to obtain resting internal

 

noise levels, the input of the compressor was terminated
with a 600 ohm resistor and the output was terminated with

a Bruel and Kjaer Voltmeter. Resting voltages were read
under each compression ratio. Results of these measurements
were -57.5 dB for the one-to-one ratio, -MM.0 dB for the

two-to—one ratio, and —33.25 dB for the three-to-one ratio.

Preparation of the CNC Test Stimuli
A diagram of the equipment used in compressing the

stimulus words is shown in Figure 8.

 

 

 

 

 

 

 

 

 

 

 

 

 

Ampex AG [Voltmeter—l Ampex AG
500-2 ' j - 606-2
Tape fl‘leiss Compressor [-— Tape

Recorder ‘ Recorder

Figure 8. Schematic diagram of the CNC test stimuli
recording apparatus.
After gain and compression were adjusted, both the recorded
calibration tape and the word lists were reproduced by an
Ampex AG 500-2 magnetic tape recorder. Output voltage of
the recorder was monitored by the voltmeter circuit of a
Bruel and Kjaer 2603 KierOphone Amplifier. The previously
recorded speech stimuli were then transmitted through the
Weiss Compressor. The output of the compressor was
connected to the input of a second tape recorder, an
Ampex AG 606-2, which was used to capture the calibration
tones and the actual stimulus materials for the study.
Performance of the compressor during the processing
and recording of the Speech was determined by the
calibration tape. Input-output relationships of the
instrument were determined at each ratio by recording the
first series of tones after they had passed through the
equipment array. Input-output functions of the mid-channel
(1500 Hz), for each ratio, are shown in Figure 9. These

Curves are very compatible with the values for beth the

1 Volt

Re:

dB

in

Cutput

C

 

 

l
\J.)
O

Input-output function
sisnal) for each compr

compressed under one--

J

(K), and three—to—one

1

~10

Volt

the mid—chanl.
on ratio when
ne compressio

low- and high—channels (i.e., 700 Hz and 3000 Hz) which are
presented in Appendix C.

The second part of the calibration tape was also
transmitted through the recording equipment under each
ratio. Recall that this section consisted of eleven
constant voltage (1 V, RES) pure tones. The purpose of
these tones was to determine the frequency response of
the processing and recording system under each ratio.
Voltage levels of these tones, after being converted to
relative dB values, are shown in Figure 10. These functions
indicated that the frequency reSponse was essentially flat
from #00 to @000 Hz for all three compression ratios.

The research design specified a comparative
analysis of the perception of speech with three degrees of
dynamic range reduction achieved by one-to-one, two-to-one,
and three-to-one comp ession. Specific procedures used in
obtaining these conditions included the following:

1. Appropriate gain and compression controls of

the compressor were adjusted to achieve the

ratio to be recorded.

{\3
o

The 1500 Hz segment of the calibration tape was
produced by the Amp x AG 500-2 Recorder. This
signal was monitored at l V (RES), passed
throuah the compressor for the ratio being

recorded, and transmitted to the Ampex AG 606—2

 

l7”

0
1

 

[1.9

O

‘4'.)

@000

I
3000

C)

Fwy

 

 

"'8
‘ C\
—O
O
1F“.
O
h-LF
N
m 0 W O in
I r—4 ,4
I I
Hf} (>091: (LG SSLIOe_UC)\J :83: , p

2000

(L) I
O (D
3.4
C) I:
£51 +3
+3
(0'

in

”1, .
V» men
I
L

A
F4; 7“
obs
+3
U) 0)
.5:- C.

03 O
l
C O
:2‘ +J
.,_; '
I!) O
C] 31:
(D 4-”
O
Q .
}._,/'\
{A O

v
"C
.‘C‘ C.
«i0

.,.,
hi) (I)

{L U'.‘
'r-i Q)
rCi 3-4
34 ,C:
O E
O O
.3 C)
3..

C)-
03 C
.{j 0
-2.) I

O
C:--{ +9
O I

C)
G) C.
’3 O
C. .
O M
C (1)
U: T)
C) Q
i: :3
>26
0 (1)
C U)
(D U)
:5 o
C H
C} Ci.
5-; E;
1:14 0

O
C: S40

(Liv
Q) Q)
34 0 .C'
:4 "4:1 O
'C 5-4 I
"-1 O O
:1. E; ~33

50

recorder used to capture the actual stimulus
items. At this point the sifinal was used in
adjusting the input record level at 0 VU meter
eflection.
The entire calibration tape was then produced
by the AG 500-2 recorder, transmitted through
the compressor, and recorded by the AG 606-2
recorder under each compression ratio.
Next, magnetic tape recordings of the N. U. No.
lists were produced by the Ampex AG 500—2
recorder. The 1000 Hz calibration tone at the
beainning of each tape was adjusted to l V (RES)
and applied to the compressor under the ratio
being recorded. The compressor output was, in
turn, used to adjust the record level of the
AG 606-2 recorder to 0 dB VU meter deflection.
Following these adjustments, all test lists
were cepied under each compression ratio.
The final recording procedure involved COpying
list IA (N. U. No. 6) for use in the pre- and
post-test. Following procedures outlined
above, word one through ten of this list were
copied with one-to-one compression. Words

eleven through twenty—five with two-to-one

51

compression, and words twenty-six through

fifty were copied with three-to-one compression.

 

Experimental Procedures

During a single test session, lasting approximately
two hours, each subject undertook the following tests in
the order given:

1. Bilateral pure—tone air- and bone-conduction

threshold tests.

2. Bilateral Speech reception threshold tests.

3. Speech-n ise ABLB test.

a. Speech discrimination experimental ire-test.

5. Speech discrimination experimental test lists.

6. Speech discrimination experimental post-test.

The subjects sat in a single walled IAC booth
having an ambient noise level of 54 dB (C scale) as
measured with a 2203 Bruel and Kjaer sound level meter and
an associated @132 condenser micrOphone. All test
materials were presented to the subjects via an Allison
clinical audiometer (22 B) and associated TDH—39-lOZ
earphones mounted in NX—hl/AR cushions.

Pure-Tone Threshold Tests. Air-conduction pure-
tone thresholds were measured by the Revised Hughson—

Westlake Ascending Technique described by Carhart and

Jerqer (1959).

a

est frequencies were 250 Hz, 500 Hz,
1000 Hz, 2000 Hz, #000 Hz, and 8000 Hz.

Bone-conduction thresholds were determined at
octave intervals from 250—b000 Hz by the Hood Technique
(1960) employing narrow band masking. The masking asent
was produced by the Allison 22-3 audiometer and its
accompanying narrow-band masking generator (Kodel 26).

Analysis of this system indicated apprOpriate critical

band widths. By applying the critical band data of

*TJ

letcher (lQMO) in the manner described by Sanders and
Rintelmann (196k), the effective masking for a zero dB
hearing level was determined at each band.

Performance of the pure-tone air—conduction system

i

was cnecked on all days that subjects were teste . A

C;

Bruel and Kja,r sound level meter (2304) and an artificial
ear (#152) were used for monitoring this system. A Beltone
Artificial Mastoid (M5A) and voltmeter contained as an
integral part of the Bruel and Kjaer 2107 Frequency
Analyzer were used to check performance of the bone-
conduction system. The calibration of these systems
remained stable throujhout the time the study was being
conducted.

Speech Reception Threshold Tests. Reception

 

thresholds for speech were measured with recorded Spondee

words. The subjects listened to the recorded test list at a

53

24 dB SL to become familiar with the individual words

(Tillman and Jerder, 1959). After this the followina

instructions were read:

You will now hear the same words aeain. At the
beginnina they will be loud, however, eventually they
will become very faint. Your task is to repeat as
many of the words as you possibly can. Even though the
word may be faint if you think you hear it, repeat it.
Do you have any questions?

Thresholds were measured by the method described by

Tillman and Carhart (1966). Specific procedures used in

obtaining these thresholds included the following:

1.

Two test words were presented at a level
approximately 30 dB above the suSpected SET.
The intensity wis decreased in 10 dB steps
With two words per level until no reseonse was
obtained for either word.

Next, the intensity was inc_eased 10 dB and
pairs of words were presented in descending
steps of 2 dB.

This process was repeated until the subject
either failed to reSpond or he reseonded
incorrectly to at least five out of six
consecutive test words.

Speech threshold was that point where the

subject last correctly identified both of th

51+

words at a level minus 1 dB for each correct
response below this point.
Broad-band thermal maskins was used during all
speech testing at the impaired ear. Effective levels
were computed and routinely applied to the good ear to
shift the non-test ear to a hearing level 20 dB lower than
the hearing level of the Speech signal of the impaired ear.
This assured that the test speech sisnal did not cross-
over to the non-test ear.
The speech circuitry of the Allison audiometer was
used to amplify and attenuate the electrical output of the
audiometer's tape deck (Viking, Model 87) used to present
all of the recorded Speech materials. This circuitry was
calibrated so that zero hearing level was 22 dB above
0.0002 dyne/cmz. Speech noise was used for monitoring
this system according to procedures outlined by Tillman,
Johnson, and Olsen (1966). Monitoring procedures for the
TDP-39 earphone include these steps:
1. The phone is coupled to the condenser
microphone (Bruel and Kjaer, Type ulna) of
the precision sound level meter (Bruel and
Kjaer, 2304) by means of a standard 6-cc
artificial ear (Bruel and Kjaer, #152).

2. The level of the Speech noise at a given

attenuator settinr is adjusted until it produces

kn
‘JT

a deflection to zero on the VU meter with the
sound level meter set on the linear scale.

3. The resulting output of the system was measured
and the value is accepted as the intensity of
the Spondee words at the same attenuator
setting under the condition in which the peaks
of the words also produced a deflection to
zero on the VU meter of the audiometer. For
example, with the attenuator set at 60 dB
hearing level, the output of the artificial
ear would be 82 dB SPL.

Speech Noise ABIB Tes . The Speech noise ABLE

 

test was administered next. The intensity of the speech
noise was fixed at sensation levels of 20 dB, 40 dB, and
60 dB at the good ear. At each sensation level the
subjects balanced the intensity of the Speech noise at
their impaired ears by a hand-held switch. This switch
controlled gain and attenuation at rates of 2 dB per
second. Five loudness-balances, with alternating
ascending and descending presentations, were obtained at
each sensation level.
Instructions given for the Speech noise AELB

test were as follows:

In this test you will hear noise alternating between

your ears. The loudness of the noise will be fixed at
your sood ear and your task is to adjust the loudness

kn
O\

of the noise in your impaired ear so that it is
equally loud with the noise in your mood ear. To
make the noise in your impaired ear softer press the
button marked 'softer' and to make it louder press the
button marked 'louder'. The loudness of the noise
will not change unless you press one of the buttons.
Allow some time for the noise to become louder or
softer. Remember that you are to make the noises
equally loud. Take all the time you need for this
balancina and make them as precise as possible. Do
you have any questions?

For practice,1fluasubjects were required to make
ascending and descending balances before the actual
testing was initiated.

Experimental Battery. This array consisted of
eight Speech discrimination tests. These included two
tests at each compression ratio and a pre- and post-
experimental test. All tests were presented at the
subject's impaired ear at 24 dB sensation level. Two
aspects of the speech material Should be recalled. First,
each compression ratio was represented in the pre- and
post-eXperimental test. Second, for the three compression
.ratios, the stimulus words were equated at their peak
:powers and the long-term dynamic ranges below these levels
ciiffered. The dynamic range of the uncompressed Speech,
zibout 24 dB, was reduced to approximately 12 dB under two-
txr—one compression and 8 dB under three-to-one.

Each subject was assigned a sequence of the three

connaression ratios. The Six possible orders of presentation

57

of the three compression ratios was used equally often.
Next, two different test lists were randomly assigned to
each of the compression ratios. A complete schedule was
constructed by first counter—ordering the ratios, and then
two test lists were assisned to each level. The following
conditions were imposed on this assignment: (1) that one
list from lists IIA-IIC and one list from IIIA-IIIC be
used with each compression ratio; and (2) that the

lists be used only once per subject. The eighteen subjects
in each recruitment group were assigned to each of these
predetermined programs.

Immediately before the pre-experimental test was
presented to the subjects, the following instructions were
given:

In this test you hear words preceded by the phrase
"say the word . . .". The words will be sufficiently
loud for you to hear them. Please repeat only the
last word of the phrase. If you think you hear a
word, but you are not sure, go ahead and repeat what
you think it might be. Do you have any questions?

Next, the subjects listened to the Six lists under
the three compression conditions after which they heard
the post-experimental test. The stimulus words were
repeated orally and the experimenter recorded their
responses on an answer Sheet. Each word counted two

percentage points. Accordingly, the percentage of the

fifty words correctly repeated was the discrimination

score for that list. The two scores for each compression

ratio were averased, thus providins the dependent vari bles

Q.)

of the study.

$31ipihizii?\

Speech discrimination performance was assessed in
thirty-six subjects for CNC words with two defirees of
amplitude compression compared with no compression.

Speech reception thresholds for CID W-l Spondee
words were used in determining the presentation levels for
the compressed Speech. Northwestern University Auditory
Test No. 6 lists were used to assess Speech discr'mination
under the three compression conditions. Speech discrimina—
tion was also measured before and after the experiment by
a pre- and post-experimental test.

The subjects had unilateral sensorineural hearins
loss with concomitant loudness recruitment in their
affected, i.e. test, ear. Recruitment was quantified as
either partial or complete by an ABLE technique that
employed Speech noise Signals. Speech noise Signals were
used because the literature indicates that subjects can

adequately balance these signals, because of their

(“(3

consanguinity with the dependent variable, and because 0
their electical nature in showing the loudness growth

function over a relativelj broad auditorv area.
3 u

\J‘K
\O

Amplitude compression was obtained with an
instrument with input—output ratios of two-to—one and
three-to-one. The system used also amplified linearly, a
condition which was called one-to—onc amplitude compression.
The compressor had 5.4 and 2.5 millisecond rise and decay
transients. Averame percent distortion for the second and
third harmonics was 2.77, 4.72, and 4.25 for the one-, two—,
and three-to-one ratios respectively.

The test battery included pure-tone air-and bone—
conduction threshold tests, speech reception threshold
tests, speech noise ABLE test, six eXperimental CNC
discrimination lists, and pre- and post-experimental CNC
discrimination tests. In the research battery, two test
lists were presented under each compression ratio with the
peak components equated at 24 dB sensation level for each
condition. Pre- and post-experimental tests acquainted the
subjects with the test conditions and provided a base level
for assessing initial group differences.

ReSponses of the subjects to the CNC words under
the three compression conditions provided the dependent
variables of the study. These were analyzed statistically
to answer the experimental questions. The results,
together with appropriate discussion, constitute the

following chapter.

CHAPTER IV

RESULTS AND DISCUSSION

‘he purpose of this study was to examine the
relationship between loudness recruitment and the
perception of amplitude compressed speech. Thirty—six
subjects with unilateral sensorineural hearing loss were
divided into two equal groups of "partial" and "complete"
recruitment on the basis of their performance on a Speech
noise ABLE test. They were given (1) a pre-experimental
test that included CNC words at three compression levels;
(2) discrimination tests under conditions of compression
and no compression; and (3) a post-experimental test
which was a repeat of the pre-test. Results of the pre-
and post—experimental tests were analyzed to determine
practice and time dependent influences.

The speech discrimination test scores for the
three compression ratios were analyzed to answer the
questions of interest which were as follows:

1. Does amplitude compression improve Speech
discrimination in persons who exhibit loudness
recruitment?

2. Is the amount of improvement in Speech
discrimination related to the amount of the

loudness recruitment?

60

61

3. Is the amount of improvement in Speech
discrimination related to the dearee of
compression?

Analysis of Practice Effects. Fre- and post—

 

experimental scores were compared to determine whether
performance changed during the experiment. Analysis of
difference scores (post-test minus the pre-test) revealed

that the subjects in each group had sianificantly hifiher

post-experimental scores than pre-experimental scores

The mean pre- and post-experimental test scores

are shown in Table 3.

 

 

 

 

Table 3. Lean percent correct discrimination scores for
the pre-exnerimental and the post—experimental tests for
the partial and the complete recruitment froups.
Recruitmentfi
Group Pre-Test Post-Test Difference t
Fartial 57.78 62.11 @033 3025**
Complete 71.44 75.11 3.66 2.77*
* p 0.05 “t p 0.0l

Althoujh discrimination performance improved, the
absolute marnitude of the rain was rather small. The fact
that both the partial and the complete groups had similar
improvement is supported by the mean rain values shown for

each sroup.

r

”"3

The question of change in

“d
(0

,rxormance during the
experimental test session, due to familiarity with the
response set, was also studied. Recall that list II and
list III of N. U. No. 6 were each permuted twice to provide
the test lists. Although the order of words was changed,
the same word sets were repeated for each compression ratio.
The mean discrimination scores for each successive order
(i.e., the first, second, and third presentations) were
averaaed over all subjects. These averages are Shown in
Table 4.

Table 4. Nean percent correct discrimination scores for

the thirty-six subjects for the first, second, and third
presentation orders for the one-, two—, and three-to-one

compression conditions.

 

 

 

 

Compression Iresentation Crders

Ratios First Second Third
One-to—one 58.00 65.25 65.75
Two-to-one 73.83 71.17 76.17
Three-to-one 72.42 78.67 75.42
Combined 68.08 71.69 72.44

 

In obtaining the values shown in Table 4, each
compression ratio was represented equally often in each
position. For example, the ratios one—, two-, and three—
to-one were each presented first to twelve subjects in

each recruitment group. The difference between the first

and third presentation averased 4.1 percent for the entire
experimental pepulation. There was no systematic trend in
performance within the three compression ratios from the
first to the second to the third presentation orders;
however mean performance was consistantly better for the
third presentation than for the first.

Analysis of Effects of Amplitude Compression. Fisure

 

11 shows the mean discrimination scores of the two
experimental groups for the three compression ratios as well
as the pre- and post-test scores. Both groups appear to
benefit from the compression. The largest increase in
performance occurs between the one-to-one and the two-to-
one compression ratios. Oddly enough, the complete
recruitment groups performance was superior throughout.

Experimental group differences were significant at
both the pre-experimental test (t of 2.29, 17 df) and the
post-experimental test (t of 2.57, 17 df). Obviously, any
comparison between the groups performance would be biased
by the difference observed at the pre-test. Specifically,
statistical tests of aroup differences must account for the
influence in prior ability.

A method is available for testing the Significance
among means which have been influenced by such confounding
variables. This procedure, analysis of covariance, adjusts

the means for the effect f the confounding variable

0

90 4

 

o \
c0 4 . \

70 C“"---0

Percent Correct Discrimination

 

 

\‘P
(:30 u
l J l ' - C
Pre— On ~to Two-to~ Three-to rost—
Test one one one Test

Fisure ll. lean sneech discrimination scores of the "Partial"
~Troun (1'8) and the "00m? ete" VTOUD (C'S) for the exocri-
e on ratios as well as the nre— and poet—

mental comnr ssi
exnerimental tests.

65

(prior ability) and makes the necessary modifications in
sampling error. The corrected sampling error is then used
to test for the differences in the adjusted means (Downie
and Heath, p. 186). The covariance analysis was used to
determine whether there was (1) a main effect due to the
two recruitment groups; (2) a main effect due to the three
compression ratios; and (3) interaction betWeen the
compression ratios and the recruitment groups.

In addition to the ere-experimental test differences
observed for the recruitment groups, the covariance
analysis also was indicated by the pre-eXperimental test
being correlated hifihly with the three compression ratios

0
F:

(r values of 0.86, O.cc, and 0.84 for one-, two-, and three—

to-one compression ratios reSpectively). The design fits a

2 X 3 fixed model, with repeated measures on the compression

variable. Factor R represents the two recruitment aroups

and factor C represents the three compression ratios. The

covariate measure (the rre-exeerimental test score) for

all criterion measures for factor C is constant for each

subject. Accordingly, no adjustment is required for the

main effect because of the compression factor (Winer, p. 614).
Results of the analysis of covariance are summarized

in Table 5.

00

Table 5. Summary of analysis of covariance comparine
performance of the two recruitment sroups for one-, two-,
and three-to-one amplitude compression.

 

 

 

Source of Variance df NS E ratio
R (adjusted) 1 228.76 1.M@
Subj. W. R (adjusted) 33 159.27

C 2 1584.45 37.80**
RC 2 28.12 0.67
Residual 68 hl.9l

 

** p 0.01

There is no sienificant effect for recruitment,
nor is there any evidence for an interaction between
recruitment and the compression ratios. The analysis
indicated that there is a sienificant main effect due to
the compression ratios. The om ibus covariant analysis

s not, however, provide specific information relative

DJ

0

I l
\

to the differences between the individual means. Rather,
it only indicates whether or not a significant difference
exists or does not exist for a particular factor. Therefore,
in order to further evaluate the significance of the
compression conditions the data were subjected to multiple
comparisons using Duncan's New Fultiple Range Test (Edwards,
p. 136-157). These results are shown in Table 6.

This analysis reveals the following: (1) the two-
t0~one mean was sianificantly different from the one-to-one

mean; (2) the three-to-one mean was sienificantly different

67

Table 6. Duncan's new multiple range test applied to the
differences between the three compression ratios. (N=36).

 

 

Compression Ratios

 

 

one-to-one two-to-one three-to-one
Means 63.14 73.72 75.36
one-to-one 10.58%* l2.22**
two-to-one 1-64

three-to—one

 

**p 0.01

from the one—to-one mean; and (3) the two-to-one and the

three-to-one means do not differ significantly.

fl

Supplemental Analysis of the Ezfects
of mplitude Compression

Careful inSpection of the speech noise balance
data, which was used in grouping the subjects, revealed
that the adopted system may not be the most suitable one

for quantifyini loudness recruitment. Specif

Ho
O

90

H
|._.l
P.
d-

was observed that subjects who appeared to have comparable
arowth in their loudness functions were classified not

on the basis of their recruitment but on the basis of the
magnitude of the interaural difference in their hearing
levels. This is illustrated by the two balances shown

in Figure 12.

/ i

{J,\(

 

c0

~18 0 (threshold) 0 (thre.
10
20
30
11,0
50
60 Q__,
70
80
CO X
100 3

a
;
u
A

,._J

old)

 

Nix:
N4

H
b.)
O

 

 

Subject A Subject 3
"complete recruitment" "partial recruitment

I 1

A318 balance data of two

(I
J
m
0
O
;:
l
:5
O
[J
m
m

Fiaure 12. h
experimental subjects.

Recall that the points (Figure 12) for the hood ear

(i.e., the 0's) are at 20 dB intervals in sensation

level and the points for the poor ear (X's) are plotted as
a mean of five equal loudness balances. As shown by the
fisure, the suojects exhibit similar relative incremenns
in judsed loudness (X1, X2, and X3) correspondins to the
20 dB, 40 dB, and 60 dB sensation levels presented in
their aood ears. For example, the chanae i the poor

ear from X1 to X2 is 5 dB in both cases as the sensation
level changes from 20 to 40 dB sensation level in the good
ear. Furthermore, both poor ear balances covered a range

to X3). This indicates that the growth 0:

O
H)
H
U1
Q)
m
V:
[.3

,—

-._, o ' _‘ -" o . o " 0"“- r ‘
'nese suogects loudness functions lS very similar.

Despite this fact, the orisinally adopted classification

(S 0

J)
'3
:3“
”D
1
)
CO
J
H
{D
O
D
Q.)
d‘
’1‘.
3
P
’3
Hr
0
Z)
H
H~
1.6
”D
t“
O
3
,—*—

aroups. This susfiested
that an alternate system for catesorizins the subjects'
loudness balance data mL ;:ht uncover some relationships
be t en recruit~e nt and compression that were Mb cure d
by the earlier classification system.

The traditional system for classifvi-‘ loudness
recruitment is contaminated, in some cases, by the
masnitude of interaur 1 difference in hearins level. In

order to circumvent this contamination, an index of

’1

‘

ec L. itment (Ir ) \as deveIOped tha.t estimated the growth

of the function in the impaired ear. To obtain the

subject's Ir, the Frowth of loudness in the impaired ear

4.

(i.e., X3- X1) is subtracted from the total ranse of

U)

ensation level presented at the mood ear (i.e., 40 dB).
For the sense ation levels of this experiment, the Ir

fornula can be stated:

:: LL - - V
II‘ .0 (X3 [Kl
Ir scores can vary from a score of 0 for subjects

who show no interaural difference in their loudness
function growth to a score of 40 for those subjects who
balance all the sensation levels presented at the good

ear with a sinele level in their poor ear. In other words,
subjects who show no recruitment receive a score of 0,
while those who show more recruitment receive higher

scores. Figure 12 provi dcs convenient data for illustrating

7O

scoring procedures. Both subjects show loudness srowth
ranqes of 15 dB in their impaired ear. These values

subtracted from the 40 dB constant yield Ir scores of 25

for each subject. Similarly, I scores were computed for

r
all thirty-six subjects. The frequency distribution of
these scores is shown in Figure 13.

The histosram (Figure 13) shows that the subjects
distribute themselves throushout the range of possible
scores with approximately half above (n of 16) and half
below (n of 20) an Ir score of 20. Accordingly, the
subjects were rearouped into two new recruitment groups
according to their Ir score for the Speech noise ABLE
test. Subjects with Ir scores of 20 or less were assigned
to Class I recruitment. Class II recruitment consisted
of subjects who have Ir scores greater than 20. In other
words, the two classes of recruitment fall on a continuum
of 40 dB (re: growth in loudness) whereby Class I
represents less recruitment than does Class II.

The pre- and post—eXperimental tests and the
experimental Speech discrimination data of the two new
recruitment sroups were again compared to further
investigate the relationship between recruitment and

amplitude compression. Mean speech scores of the two

groups are plotted in Fifiure l4.

1. 5 a
d
1
7“}
.3 "1
o .1
O O
.r: _
Q 1..
S -
U)
-
“H
0 d
S—I _
(1)
r9 Bu
.:3_ d \ fi l
d
-
l

 

 

71

 

 

 

 

 

 

Fisure 13. s
recruitment (Ir) scores

subjects.

,
, £1

nistoaram

T the sneech nOise incex o:
the thirty-Six experimental

"\I
i\)

H

0:3 a"

c 90 j
d

O 1-------

mg-

 

 

+J ~
s . “~qI
G) CO d
O
a
(D
I I 1 1 I
Pre- Cne-to- Two-to- Three—to— Eost-
Test one one one Test
“idnie lb. loan sneech discrimir” ion scorcs of the L ear
Class I (I's) and the sic'een Class II (II's) subjects for
the excerimental CO“1fcs ion ratios as well as the ore- and
nest—evrerinental te3ts. (‘las s I renresents less
recruitment the n Class II).

73

When classified by the Ir’ both sroups asain ap.

w

to benefit from the amplitude compression. The larses t

gain in performance also occurred between the one—to-one

and the two- to- -one ra atio. The groups did not differ at

the pre—experimental test (t of 0.4M, 1? df) or at the post-
experimental test (t of 0.M3, 1? df).

In contrast to the original analysis the Class I
group performance was slightly superior throughout.

This is clearly evident when one compares Figures ll and

In. This finding demonstrates the necessity for additional
research concerning methods for quantifying the "amount" of
loudness recruitment.

An unweighted—means analysis of variance wa
utilized to determine if statistically significant
differences existed: (1) between Class I and Class II
recruitment; (2) be ween the three compression ratios; or
(3) among the classes and the compression ratios. A

3 f ixe ed model, with repeated measures on the compression
factor was utilized. Factor E represents the two
recruitment sroups or cla ms es. Factor C represents the
three compression ratios.

Results of the unweighted-means analysis are
summarized in Table 7. Differences between the recruitment

groups classified by their Ir were not significant at the

0.05 level (F of 0.43). Differences a~mong the compression

7/3

 

 

 

 

Table 7. Summary of unweishted-means analvsis of variance
comparina performance of Class I and Class II recruitment
for the one-, two—, and three-to—one compression ratios.
Source of Variation SS df MS F

R 140$}.9ZL I. 1>0<‘.019 0.193
Subj. w. R. 32520.80 3M 956.e9 __”___
C 3122.90 2 1561.b5 31.M4*
RC 1.25 2 0.63 0.01
Residual 3372.84 6? h0.70 ____~_
T*:> 0.0]-

ratios \ere found to be statistic ally si"nificant at tie

o 01 level (*P of 31.bb)

The sier ificant main ef: ect for the ceinr ession
ratios was further evaluated by multiple comparisons usinm
Duncan's New Iultiple an e Test (Edwards, pp. 136-157).
This analysis yielded identical results with those
presented in Table 6. call that the former analysis
indicated that (l) the mean of the two-to-one compression

was different from the one—to-one compression, (2) the
th ree-to-one mean was sifinificantly different from the
one-to-one mean, and (3) the three-to-one mean was not

sisni ificantly diff rent from the two-to-one mean.
Gain in speech di Cllm‘nc tion between the three
ratios is shown in Table 8.

75

Table 8. Deans, ranees and standard deviations of percent
gain in speech discrimination scores from the one-to-one
ratio to the two-to-one ratio, and from the one-to—one
ratio to the three-to-one ratio (N=36).

 

 

 

Standard
Gain Comparison Kean Deviation Range
Cne-to-one to
Two-to-one ratio 10.58 9.55 -3 to 32
One-to-one to
Three-to-one ratio 12.22 11.56 --13 to 38

 

Discrimination improved significantly between the one-to—one
and the two—to-one conditions and was also significantly
better for the three-to-one condition when compared with

the one-to-one condition.

Summary of th Results

(D

 

An experimental battery consisting f (l) a speech
discrimination pre-test, (2) discrimination tests under
conditions of compression and no compression, and (3) a
post-experimental discrimination test was administered to
thirty-six subjects who exhibited loudness recruitment
for a speech noise ABLE test. Two classification systems,
based upon estimation of the magnitude of the 10 dness
recruitment, were utilized for placing the subjects into
experimental groups. The first system involved a comparison
of the magnitude of the sensation levels above threshold

1

necessary to make the test signals equally loud. This is

76

the conventional method of quantifying recruitment
according to Jerger (1962). The second system, an index of
recruitment (Ir), was based on intra—aural comparison of
the loudness function growth.

The data were analyzed independently for both
systems of classification. Results of these analyses were
strikingly similar in that both indicated (1) no
significant main effect for the recruitment factor, (2) a
significant main effect for the compression ratios, and
(3) no significant interaction between recruitment and the
three compression ratios.

Additional evaluation of the significant main
effect for the compression ratios indicated that the two-
to-one and the three-to-one ratios were significantly
different from the one-to-one ratio but not significantly

different from each other.

Discussion of the Results
Data concerning the discrimination of amplitude
compressed speech by persons with loudness recruitment
were provided by this study. Analysis of these data
provided the basis for this section which will address,
individually, the three experimental questions.

1. Does amplitude compression improve Speech

 

H.

d scrimination in recruiting ears?

 

‘0
q

The evidence clearly indicated superior Speech
discrimination for the two-to-one and the three-to-one
compression conditions when compared with the one-to-one
condition. Average performance of all subjects was 63.14
for the one-to-one condition, whereas it improved to
73.72 and 75.36 for the two-to-one and three-to-one
conditions, reSpectively. These findings supported the
experimental hypothesis that persons with loudness
recruitment would be afforded Special benefit in their
understanding of speech signals by amplitude compression.

The observed gain in speech discrimination might
lead to the arguement that it is the dynamic range rather
than the sensation levels of the strong components which
had the greater influence on the sensorineurals' Speech
perception. Cr, stated differently, when the peak powers
of amplitude compressed speech are equated, the individual
components assume new relationships and these changes
appear important for perception. This is a very attractive
hypothesis, but only partially confirmed by the results of
the present study. A definitive answer must await furthe
evidence based on other approaches. For example, it
should be known whether comparable results would ensue by
equating the peak components of the speech signals at

other presentation levels.

‘3

The improved speech discrimination in this study
has not been observed previously. Barker (1953), lynn
(1962), and Caraway (1964) have investigated hypoacusic
perception of amplitude controlled speech. Contrary to
the findings of this study, these three reported no
systematic beneficial effects with amplitude controlled
speech. The conflicting outcomes can perhaps be best
explained by careful attention to differences in method
of amplitude control and in the fidelity of the equipment.

Parker (1953) and Lynn (1962) utilized a method of
amplitude limitation for controlling their speech signals.
The discrepancy between their results and those of this
study might well be due to differences in the dynamic
functioning of compressors and limiters. Caraway's (1964)
research, however, involved amplitude compression using
the same compressor that was utilized in this study.

It is imperative, then, to determine a reasonable
basis for the discrepancy between the results of this
eXperiment and those obtained by Caraway. The most
reasonable explanation can be found in the performance
variation of the Weiss compressor during stimuli processing
and recording during the two separate experiments. The
compressor has been modified since it was used by Caraway.
This modification provided more rapid roll-off of the

frequency response below 400 Hz and above @000 Hz.

\O

Althouqh restricting the frequency response, these changes
reduced noise and harmonic distortion. Performance
differences are evident when one compares the percent
harmonic distortion before and after the modifications.

The averages of Caraway's percentage distortion measurements
of the second and third harmonics were 4.93, 10.09, and
22.98 for the one-, two-, and three-to-one conditions
reSpectively. Corresponding measures, after the
modifications, were only 2.77, 4.72 and 4.25. An analysis
of the compressor's performance by Caraway indicated

severe harmonic distortion in the low frequencies with
systematic decrements in the distortion in subsequently
higher frequencies. Apparently, most of this distortion

was eliminated by the modifications consisting in rapid low—
frequency rejection.

The commonly held opinion that speech discrimination
suffers in the presence of harmonic distortion is documented
by several research reports. Harris et. al., (1961):
Jereer, (1967); and Kasten et. al., (1967) found positive
relationships between speech understanding and harmonic
distortion: Specifically, all observed Optimum discrimination
under conditions of minimal distortion. Thus, eXplaining
differences between the results of this study and Caraway's
results on the basis of harmonic distortion differences

appear warranted.

80

Rise-decay transients measured after the instrument

modifications are Slightly longer than the values measured

*3

F

by Caraway. She measured rise and decay transients o'
1.5 and 1.0 msec. respectively. Corresponding values,
after the modifications in the present study were 5.4 and
2.5 msec. These findings should be interpreted with
considerable caution because they may have resulted as an
artifact of measurement. Also in the event that they are
real differences, their rather small magnitude would tend
to limit any systematic effect on discrimination (Lynn, 1964).

2. is the amount of improvement in speech

discrimination related to the amount of the
loudness recruitment?

This question asks whether increase in the
intelligibility of compressed speech is related to the
amount of loudness recruitment and does not concern
absolute discrimination performance.

The thirty-six subjects involved in the study
gave evidence of loudness recruitment for a speech noise
ABIB test. Data relative to their perception of three
degrees of amplitude compressed speech were partitioned
into two groups according to their amount of loudness
recruitment. Actually, two systems were utilized for this
partitioning. The first involved comparing sensation

levels necessary to make the speech noise signals equally

El

loud. The second system compared, intra-aurally, the
loudness function erowth.

Neither of these methods for definina recruitme Mn“

offere ed evidence that amplitude compression he s a
differential effect on the two recruitment Groups. Both

the partial and complete O“roups showed similar amounts of
Cain.

The subjects who exhibited the most recruitment
and the most gain (benefit) for the compression were
studied in detail. Ir scores for the nine subjects who
experienced the larsest and the smallest rain in Speech
Mi crimination are shown in Fisure 15. The distribution
of these scores, for both low— and .iph-eain subjects,
covered a wide ranse and provide little evidence of a
relationship between the amount of recruitment and gain
from compression. A similar treatment of gain was obtained
from subjects with clearly high and low Ir scores for the
Speech noise ABLE test at the 40 d3 sensation level. Ir
scores computed for all subjects at this presentation level
revealed seven subjects with distinctly low Ir scores and
ten with distinctly high Ir scores. The seven subjects
xvith low Ir scores demonstrated average pain scores of
24.71 percent, whereas the ten high Ir subjects demonstrated
comrmtible average gain of 20.30 percent. Again, the

results definitely do not support the experimental

10—

: low—Gain Subjects

—
5d 4—
-
C”? -
P
O -
C)

 

 

 

 

Sub
l
UH
}_.l
0
Ha
\ft
‘8!
\fig
\gq:]

Ol

”99?

- a
I l ‘v’
‘\ 1|
A. .gAAI

‘
/

v

Eish-Gain Subjects

K.“

 

m-
H
O
H
UK
N-
O
to
J1
b)
0‘
Lo
C)

‘corcs

m
J
O
(L)
O
O
P o
('l
KD
}.
a
”J
W
O
' 3
w
\D
O
L}:
E.
r"
s
CD
:3
(.1.
A
H
H
V
LIN

lflirnire 1J3. E

cise index or
rweczmiitxnen. *- p

" : .'-
. C CUDJCCLS
. O Q l V
_ e . nirii-rr in) zmud ‘ohe
"“Wbllflfll 1 “"’ in 71r‘”ip fiNTWi“”11m “rvwavh’ a
gnuei... ,u L; “Ox-"5 {tut 0" 1.4.22 .1.) _ _._- 11L, (1~.;. COM.‘ J_ 2 ..:.).LO.'. .

F+
*4
J
D
c+
f
3
N
O
O” ('1‘
0
D |-
0 O
} I
3-
(._'I
PJ—

n

3

hypothesis that compression would have a differential
effect on the two recruitment groups.

3. Is the amount of improvement in speech
discrimination related to the desiee of the
compression?

This question asks whether there is a systematic
effect produced by the amount of amplitude compression.
The subjects responded to speech discrimination tests
delivered under three compression conditions. These were
rather widely Spaced discrete points along a continuum of
conditions which can be specified by their input—to—output
ratios (in dB). linear amplification bounds the lower
limit of the continuum and, as such, is specified by one-
to—one compression. A two-to-one ratio represents moderate
compression, whereas three-to-one is rather extreme
compression.

Data relative to the influence of the compression
ratios indicated a significant main effect for the
compression conditions. Subsequent analyses indicated
that averafe performance for the two-to-one and the three-
to—one ratios were significantly better than the one-to-
one ratio but not different from each other. These
findinss do not support the experimental question statins
that the amount of improvement in speech understanding

\vould be related to the amount of compression.

The evidence does not suggest that compression did
not enhance the subject's speech discrimination. Rather,
it simply implies that no linear or systematic relationship
was found between the amount of discrimination enhancement
and the degree of compression. This relationship is
evident by the amount of gain in discrimination realized
from each compression condition. Average gain for all
subjects between the one-to-one and the two-to-one ratios
was 10.53 percent. However, the average improvement
between the two-to-one and the three—to—one ratio was
only 1.64 percent. In other words, the evidence indicates
that the subjects were benefited about equally for the
two- and three-to-one ratios.

Range and standard deviation values of the gain
scores (see Table 6, p. 67) indicate very large subject
variability. Only three subjects failed to benefit from
the two-to-one condition and only four failed to benefit

from the three-to-one condition.

Summary and Clinical Implications
Spmmarv. Within the confines of the study, three
tentative conclusions can be advanced. First, amplitude
compression appears to enhance speech perception in persons

Vmith loudness recruitment. Second, the amount of

enhancement does not appear to be related to the degree of

C I
‘J\

the recruitment; and third, neither does it seem to be
related to the amount of compression. These findinfis have
obvious implications; however, certain ther secondary
findings emerged which may have important implications
resardins our clinical management of hypoaeusis.

Recruitment and the speech noise All: vest. here

than sixty persons with unilateral sensorineural hearing

lo s were examined in order to obtain the specified

(0

experimental sample. failure in meeting the pure tone
preliminary specifications was the predominant attrition
factor. most of the subjects could adequately perform

the loudness balances; however, many needed repeated
instructions and encourarement. Two peeple were encountered
who simply could not perform this task.

Only two subjects failed to exhibit loudness
recruitment for the speech noise sifnals. To check this
findine, ABLE data for pure tone sifinals (i.e., 500 Hz,
1000 as, and 2000 Hz) were obtained for all subjects. The
speech noise and the pure tone signals yielded compatible
results. Recruitment was indicated by the speech noise in
every case where it was indicated for a single pure tone
or~2for any combination of the pure tones. The average Ir
for-the three tones and the Ir for the Speech noise
correlated at 0.MQ. “inally, the two signals used for

ILLLB (pure tone and speech noise) yielded compatible

CO
O‘\

results in a small control sample (N of 6) of persons with
conductive hearing loss.

These findings have two important clinical
implications. First, if the experimental sample is
representative, one can expect some degree of loudness
recruitment in practically all subjects with unilateral
sensorineural hearing loss. Second, speech noise appears
to be a good ABLB test signal.

External Validity of the Results. Several factors
prohibit generalizing the findings of this study to
wearable amplification (i.e., hearing aids) utilizing
amplitude compression. The CNC signals were used because
they enabled quantification of the speech perception.
vaiously, the dynamic range of these signals was not
representative of everyday speech and the subjects listened
in an artificially quiet environment. Koreover, the Weiss
compression amplifier is an AC Operated desk—type
instrument and, as such, cannot be worn on the body.

Clinical Implementation of Amplitude Compression.

People with sensorineural hearing loss generally
exhibit loudness recruitment, and they discriminate speech
better when it's amplitude dimension is compressed. Some
few do not experience this benefit, whereas others
exhibit dramatic improvement. Although the benefit of the

copmmession cannot be predicted by the recruitment, simple

clinical procedures could identify these people who
potentially will be helped and could estimate the magnitude
of the help.

Since there is no relationship between the amount
of benefit and the amount of compression, this identification
could involve a simple comparison between discrimination
for compressed and uncompressed Speech signals. People
who are substantially helped should have great potential
for success in utilizing amplification employing amplitude

compression.

CIA? T7313. V

SJEYARY, COKCLUSICNS, AND RECCIKZTUATICNS

Speech discrimination performance of thirty-six
subjects was compared for conditions of amplitude
compression and for no compression. These conditions were
achieved with a prototype amplitude compressor built by
Kr. Erwin Weiss and supplied by the Beltone Electronics
Company. The instrument had in put- to- output amplification
ratios of two-to—one and three—to-one, and it also
functioned as a linear amplifier which was called one-to—

D
I

one ampli ic ation.

Th

(D

thirty-six subjects, all of whom had unilateral
sensori neural heari ins lo oss, were partitioned into partial

or complete loudness recruitment groups on the basis of
their response to a Speech noise ABLE test. Cla sifi cation
of subjects was carried out accordine to two procedures.

The "classical" procedure involved a comparison of the
magnitude of the sensation levels above threshold necessary
to make the test signals equally loud. The second system
was based on intra-aural comparison of the loudness function

growth.

{“0

Q-)

c AU
\0

CNC test material, under each compression condition,
was presented at 24 dB sensation level at the effected

ear. Broad band masking noise was routinely applied at

'1)

f the data,

0

inaly‘es

U"

the contralateral ear. Statistical
partitioned by either method for defining recruitment,
yielded similar results, namely: (1) no siinificant main
effect for the recruitment factor; (2) a significant main
effect for the compression ratios; and (3) no significant
interaction between level of recruitment and three
compression ratios. Additional evaluation of the
significant main effect for the compression ratios indicated
hat discrimination scores for the two-to~one and the
hree-to—one ratios were significantly higher than the
one-to—one ratio but not significantly different from each

other.

Conclusions

 

Within the limitations of the experimental design
of this study and the instrumentation employed, the
following conclusions are warranted:

l. Amplitude compression enhances speech

discrimination in persons who have unilateral
sensorineural hearing loss with concomitant

loudness recruitment.

\C)
O

2. The amount of discrimination enhancement is
not related to the magnitude of the loudness
recruitment.

3. There is no relationship between the discrimin—
ation enhancement and the maenitude of the
amplitude compression (two-to—one versus
three-to-one ratios).

M. Speech noise signals can serve as adequate
ABLE test stimuli.

5. loudness recruitment lS exhibited by most

persons with unilateral sensorineural hearing

6. The omnibus aural rehabilitation process should
include assessment of discrimination enhancement
provided the hypoacusic by compression
amplification.

7. lersons who demonstrate superior performance
for controlled speech sirnals with compression
amplification may have great potential success

in utilizing individual amplification with

ii ‘ o
necommencations

 

Although this study found that hypoacusics are

‘benefitted substantially by amplitude compression, many

91

questions remain to be answered. huch work needs to be
done regarding discrimination of amplitude compressed
speech in persons with cochlear problems.

Since no attempt was made to classify and
systematically study the etiology of the hearing loss,
such an investigation should be undertaken. Discrimination
of compressed Speech, like discrimination for distorted
Speech per se, may be quite sensitive to the type of
end—organ lesion.

The effects of noise masking on understanding of
compressed Speech in sensorineural subjects should be
studied. Differential effects of various types of cochlear
lesions in identifying noise masked compressed speech
should be researched.

Nultifactor studies should be designed to
systematically investigate enhancement of speech perception
in end-organ lesions. Such research should include
Specification of the pathological ear's performance with
respect to time and frequency parameters. Such research
also should study psychological variables such as
preference for compressed Speech versus enhancement in
Speech understanding.

The present study should be replicated using a
variety of speech signals presented at numerous sensation

levels. For reliability, a replication using N. U. No. 6

\o
:o

is needed. This should be followed by investifiations with
a variety of stimulus materials such as The Fairbanks'
Lodified Mhyme Test or a sentence intelligibility test.
Finally, much work remains to be done relative to
our clinical implementation of amplitude controlling
devices. Data about the performance characteristics of
hearing aids currently available with some form of
amplitude control should be obtained. The Weiss instrument
should be compared with a hearing aid that is capable of
achieving compatible control. Objective and subjective
data concerning the effects of this form of amplification

in a wide variety of applications is needed.

BIBLICGRA. m

BIBLIOGRAIHY

Ancona, E. P., Jr., Compression and dialog equa aliz ation in
motion picture and sound recording. Aup io, 40,
17-19 (1959)-

-_-_-——'—-.—c

Ashton, J. P., Desisn Of Commercial hearing aids. Jouzmgl
itish Institution 0: F;a(iio Engineers, ll,

AsDinall, E., Desisn of commercial hearins aids. uouinal
of the :r- tish Institution of Eadie jaci‘oc1., ll,

E5"50 (1751)-

U)
H.
N
C)
'3?

Lu

(D

t, G. and Veit, P.,
deaf children. Diffi
larynsol. Ctol. Hhi o

dv vanced apparatus on
'is and results. heview of

 

grander, R., (Beltone Electronics Corporation ,Chicago,
Illinois), Eersonal Communication, Em ay 1970.

U]

Caraway, B. J., " ffects on Speech Inte llisibility of
Dynamic Rm mje Reduction with a Compression
Amplifier." unpublished lh.D. Dissertation,
Northwestern University (196M).

J. and Cawrh art, 3., Influence of compressor

 

Caraway, 3.
action on Speech intelll"lbllltV. J. Acous. Soc.
AL'er., Ml, lWZN-lJIBB (1967).

Carhart, R. and Jerter, J. F., A preferred method for the
clinical determination of pure tone thresholds,

 

 

J. Speech hearing Dis., 2%, 30- 3M5 (11959).
Clemis, J. D. and Carv ver, W. F., Discrimination scores for
Speech in Keniere's Disease. Arch. Ctolaryn7., 86,
O
614-6lo (1967).
Davis, S., Stevens, S. S., Nichols, R. H., Jr., Hudains,

7‘1

C. U., Larquis, R. J., Peterson, G. n., and Ross,
D. A., chrio" Aids - An Experimental Study of
Decisn C'b woerlves, Cambridge: Harva1d Lniversi ty
lress (1997).

 

 

93

7'

Davis, h. and Silverman, S. R., He;r kn and DeatrOJS. New
s

ton, Inc. (1970).

 

York: Holt, Rinehart and U n

more:

‘Jc

 

Denes, P. 3. and Pinson, E. D., The Speech Chain. Salt
Waverly Press, Inc. (19663.

Dix, N.R., Hallpike, C. S. and Hood, J. D., Coservations
upon the loudness recruitment phenomenon, with
especial reIeIence to t? e diIf eIentia.l diarnosis
of disorders of the internal ea r and VIIIth nerve.
J. Iaryntolo:y Ctology, 62, 671—656 (1948 .

 

Downie, N. N. and Heath, R, W., ic S+°tl"IlC”l Aethods.
New York: Harper and Ron %(l965).

 

Eby, I. G. and Williams, H. 1., Recruitment of loudness in
the differential dia nosis of end- or an and nerve
fibre dea.I ness. laryngoscone, 61, LOO- 914 (1951).

 

Edsardh, B. J., The use of extreme limitation for the
dynamic equa1i7ation of vowels and consonants in
hearing aids. Acta Ctolaryns. , #0, 376-386 (1951).

Edwards, A. L., Exnerimental Desifin in Psychological
Deeearcu. New York Holt, Rinehart and Winston
(19'6” 0T.

 

Flemminr, D. B. iand Rice, C. G., New develonment concepts

 

 

In he fraids. Int. Audiol., 8, 517-523 (1969).
Fletcher, H., Auditory patterns. Rev. Kodern Physics, 12,
47 5 (19b00)o
Fournier, J. E., Audiometrie Voc 19. Paris: Iibrairie

 

Naloine s. A. (1951).

Grimwood, W'. K., Volume compreSSOIs for sound recordins.
J. Soc. lotion PictUre En rs., 52, 49- 76 (l9br9).

 

Groen, J. J., Aanmeten van eehoortoestellen (Hearing aids
made to order). Ned. Tijdschr. Geneesk, 158 (1951).
k>ce rnta Nedica, 4, 1239 (1951).

 

 

Custavsson, L., Speech Amplitude CO'nIeosion Svstem of the
Narcou-Daguet Tyre. SEI-gron-l/1965. Stockholm,
Sweden: Speech Transmission Lab., Royal Inst.
Technol., 14 (1963).

 

 

, J. D., Haines, H. L. and lyers, C. K., loudness
nercebtion for pure tones and for sneech. ggch.
Ctolarrn”.. 55. 107-133 (1952).

 

Harris, J., Haines, H., Kelsey, P., and Clack, T., The
relation between speech intelligibility and the
electro-acoustic characteristics of low fidelity
circuitry. J. Aud. Res., 1, 357-381 (1961).

airsh, I. J., Davis, H., Silverman, R., Reynolds, E. G.,
Eldert, E., and Benson, R. W., Develonment of

materials for sneech audiometry. J. Sneech Hearing
Qiéis 17. 321-337 (1952).

 

Hood, J. D., Principles and practice of bone conduction
audiometry. larynroscone, 70, 1211-1228 (1960).

 

Hudgins, C. V., Narquis, R. J., Nichols, R. H., Jr.,
Peterson, G. E., and Ross, D. A., The oomcarative
oerformance of an exnerimental hearing aid and two
commercial instruments. J. acous. Soc. Amer., 20,
2u1-258 (1948).

 

Huizins, H. C., Symctoms of recruitment and sneech
intelligibility. Acta. Cto-larynsol., 36, 346-355

 

Huizins, H. C., The recruitment factor in hearing tests.
Acta. Oto-larynsol., #0, 297-306 (1952).

 

 

Huizing, H. C. and Reyntjes, J. A., Recruitment and sneech
discrimination loss. laryngoscone, 62, 521-527
(1952).

Jerger, J. and Harford, E., Alternate and simultaneous
b'naural balancing of pure tones. J. Speech Hearing

 

 

Jerge

w

, J., Hearing tests in otologic diagnosis. Asha, 4,

1939-1M5 (1962).

Jerger, J., Behavioral correlation of hearinn-aid
performance. Bu11. Prosthetics Res., BER 10-7, 62-

75. (1967).

 

Kasten, R., Lotterman, S., and Burnett, 3., "The influence
of nonlinear distortion on hearina aid processed
sisnals," Pacer nresented at 1967 convention of the

merican Snecch and Hearing Association.

\“etsinser, E. A. and Younfi, N. 3., The Use of fast limitins
to iMOIOVB the intelliTiblliIY 01 S eecn 1n noise.
Snecch loncsr., 27, 63—69 (196C).

 

.i"ner's Handbook.

lansford-Smith, E., Editor, {ad' 33
alve Company Ltd.

1
Sydney: Amalfiamated Wir
(1952).

lehistc, I. and Peterson, G. E., I
in the study of sneech int
Soc. Amer., 31, 260—236 (1

 

1
11i¢.:oi1ity. J. Acous.
£31..

 

licklider, J. i. C. and Pollack, 1.. Effects of
dif :rerentia tion, intefiration, and infinite oeak
O

llrnins uron the intellieibility 01 Shgech.
J- ACOW - 306- Ancr.. 20. 42-51 (lane).

 

lynn, G., "Influence of Comnression Amnlific ation on the
Percention of Sneech by the Recri.j Tmniired
Unnublished 1 .D. dissertation, Mtiwestern
University (1962).

lynn, G. and Car*al+ R., Influence of attack and release
in comnression 8”“1111C.t10n on understanding of
sneech by hyooacusics. J. Sieech Hearinm Dis.,
23, 124-140 (1963).

Narcus, A. and Karcus, W., Elements of Radio. Enslewood
Cliffs, New Jersey: Prentice Hall, Inc. (1965).

 

rmszwe11,D. E., Dyn mic performance of nee ii limiting
amnlifiers.a 1100. Institute Radi oEngjneers. 35,

 

1399-1354 (19WAL 7)

Penzel, O. J.
Eve,

Com mn res sion amolification in hearings ids.
er, Pose, Throat Vorthly, 95, 98-101 (1966).

[15} a

V“,

Teuberfer, F. 1., Cimnression correction of hearins aid
outnut by means of interchanseable resistors.
Konatsschrift Ghren Heilk, 85/1, 1-7 (195“).
Excernta Nedica, XI, 7, 351-352 (1954).

 

Newby, H. A., Audiolosy. New York: Annleton-Century-
Crofdms (196%).

 

s of Evnosure to White Noise on
eu"”nontc and Intelli ibility

§§£gf§, Joint roject Report 29, U. S. Naval

School of Avie ”tio hedicine (195%).

_",T ‘ ‘T’ T 1. Taps
CI' _l\e]-ll’ U. do , T119 $11.1.J. C

Loudness DalanC;

| (’1‘

\D (D

*T.

97

O'Neill, J. J. cand Cyer, R. J., Anolied Audiometry.
New Yorl:: Dodd, head, and Comnany, Inc. (1966).

x
/

 

Parker, C. D., "The Effects of the Reduction of 'Short
Time Fatigue' on Seeech Intellisibility for

'Percentively' Deafened Individuals." Unnublished
Ph.D. Dissertation, State University of Iowa (1953).

Pestalozza, G., Studio comnarativo sull efficacia di
diversi dis nositive di controllo di volume delle
nrotesi acustiche. (Comnarative study of the
efficiency of various volume control devices of
hearing aids). Archives of Italian Ctolarvnf.,
?4/2 195-206 (19533. Excerpta hedica, XI, 6, 425
1953 .

Peterson, G. E. and Lehiste, 1., Revised CIIC lists for
auditory tests. J. Soeech He arin: Dis., 27, 62- 70
(1962).

Poliakoff, A., Rearins aid design--some essential qualities
of a good instrument and their practical realization.
Wireless Jorld. 56, 279—276 (1950).

 

Portmann, M. and Portmann, C., Clinical Audiomet try.
Shrinjfield, Illinois: Charles C. Thoma a.s (1961).

 

 

 

 

 

Pyron, E. O. and Williamson, F. R. Jr., Si nal Processin?
by Infinite Clinnins a d Related Technioueo. AD 600
833. Washintton, D. C.: Cffice Tech. Serv, U.S.
Dent. Comm. (1969)

Rice, C. G. and Flemmins, D. 8., Circuit develonment for
hearins aids. Inst. Sound Vibration Res., Annual
Renort, 29 (19i8).

Rintelmann, W. E., and Jetty, A. J., "Reliability of speech
discrimination testing using CNC monosyllabic words,"
Unpublished Study: Niehigan State University, 1968.

therford, C. R., Instantaneous sneech comoress or.
Electronics. 30,168-169 (1957).

Sanders, J. W. an. Rintelmann, W. E., Nasking in audiometry.
Arch. Otouaryngg, 80, 541-556 (1964).

Schultz, N. C. and Streeny, C. S., The sneecli discrimination
function in loudness rnciUWtinc ears. larynroscogg.
77, 2114—2127 (1967).

 

U)
1
,__)
_J
,._J
I
CT
L.
l
U)
Q

R. E., "Diff erenti 1a1 Sensitivit v to rnccUenc
01.nse in the Perceptively Deazcned Adult."

Unpublished Fh.D. issertation, Northwestern
University (1950).

 

 

Silvermann, S. R. and Herr Mi SOL, C. E., The National
Research Council Hearing Aid Troject, A R oar ms
Renort. Amer.mA111als of the Deaf. 96, b20— M31
(1951).

Sinzer, G. A., Performance and oneration of a new limiting
aiolifier. Audio Engineering, 34, 18-19 and 69—70
(1950).

Tillnian, T. W. and Jeréer, J. F., Some fac rs affectins
the snondee threshold in norms al-heinins subjects.
J. Sneech hearint Ros., 2, 11M1 116 19%9).

 

scriminetion utilizinq CNC monosyllabic words
(NU Auditorv Test No. . U. s. School of Aaron 803
Ned icine Technical Research, 66-55, 1-12 (19667.

Tillman, T W. -ni Carhart, R., An exnanded test for Speech
."

VI

 

 

Tillman, T. W., Johnson, R. N. and Clsen, W. G., “ arphone
versus sound field threshold sound- -nressure level
for snondee words J. Acous. Soc. Amer., 39, 125—

133 (1966 ).

Tolhurst, . G., The Effects of Signal -to-Noise Ratios and
feak Clinning Uoon a Time Accelevateo IUltiele-
phoice Intelli~1b1,1tv Te st. Chio State Res.
Pctmdation and U. S.N. School Aviation Ned, Report 83

(1059)

,.

 

 

I.)

 

4‘6

I

 

Velichkin, A. 1., Arnelitudnoe ogranicheni rechi. (Amplitude
clinving of sneech). Artstich 21., 2, 168-174
(1962).

 

Winer, B. J., Statist

APTENDIX A

GLCSSARY CF TERES

GICSSARY CF TERLS

mnlitude control involves restricting the lone-
term dynamic ranee and the beak comnonents of amnlified

sienals. Two devices for achieving this control, amplitude

f4
3
U)

compressors and amelitude limite , are Specified by the

net effect they exert on sveech signals.
Amnlitude Comnression: When the amnlitude

 

1

dimension of sneech is comnresseo, the dynamic ranfe
between the stronger amolitudes and the weaker amplitudes
is reduced reeardless of the level of the inout sifnal.
Comoressors, then, have variable outnut aain which is
inversely related to innut signal amplitude. They also
reduce maximum Dower output, but this is of secondary
imoortance to their Prime function of dynamic ranfie
reduction.

Awolitude liritinfl: Relationshins between the

 

stronfier and the weaker amnlitudes are unchansed durin:
low~innut levels when soeech is reproduced by a limiter.
However, these instruments nerform similarly to comnressors
for high-innut sianals. In other words, amnlitude limiters
simnly restrict the maximum outnut of amnlified signals.

Hearina Level:

 

This is the ratio, exnressed in decibels, of the
threshold of an ear at a specified frequency to a

\O
\ ID

100

standard reference zero level for euro-tone audiometers.
lractica 11v it is the readins in fiecibe s, on a

standard audioreter, that correstonds to the

listener's hearins Ml shold (Davis and Silver"er,

1970, r. ”98).

loudness Recruitment:

 

 

Auditory recruitment, which was given its name in 192a
by Dr. Edrund Fowler, Sr., is basically defined as an
normally rabid increase in sensitivity to loudness
the ntensity of a sound is raised above threshold

'Neill and Cyer, 1966, b. 132).

A

O L0 04

?
recruitment Classification: Two classification
systems were used for rlacins the subjects into ex_oeri:ne ntal
recruitment srouns. The first system placed the subjects
into the traditional "martial" and "complete" recruitment

cataeories by comnarina the maenitude of the sensation
level above threshold necessary to ma} e test sirnals
equally loud. The second sys 'em, an index of recrui.ment
(Ir)' placed the subjects into ana alosous Class I and
Class II srouns by intra-aural comnaris. on of the loudness
function erowth.

Sensation level: Sensa.tion level is "the pressure

 

level of the sound in decibels above (or below) its
threshold of audibility for the individual observer or
for a snecified group of individuals" (Newby, n. 12).

Sneech Noise: Soeech noise is noise which

 

conforms to the long-time averaee sneech spectrum. This
snectrum is derived by analyzins seements of runnins

T"

sneech in which every sound occurs many times. Lnersy

}_n
O
}_J

to

levels in each part of the oectrum are measured and

summe , separately, for the entire sequence. Finally, the

Q;

summe energy for each part of the Spectrum is slotted

(Denes and Finson, 1963).

AEIBNDIX B
gesrcrsas CF THE GROUPS TC

PRELIMINARY TESTING

.MopoEOstm
esp mo mPHEHH sssflxsfl mew Geomep one; mmmcoemem eEom mmsmoop UoPSQfioo so: csoge

 

msuoa.om.mm mm.mm oaanma ma.mw * ow10a.sa.em ma.mm caflnom om.ms * oooa
mmuo oo.om HH.em omumm ms. m Hw.mm osnm om.mm ma.em mcnmm ms.ms ee.ee coo:

mmuo om.eH mm.mH mono: me.om mm.mm mmum om.ea mm.uH msnmm mm.em mm.sm ooom

omum ow.mH ea.HH menmm om.m: oo.me omum sH.eH He.ma osnmm ms.mm am.em 00m

mmno mm.sH ma.efl osnmz Om.sm om.ms mmnm we.eH mm.ma meuom om.mm

 

emcmx Cc; see; macsm sea smog omssm Cs; smog owes; Ge; smog mm
new Pmmecoz gem ewes Mam Pmeesom hem ewes

mpufiw mamaqguo Fromm HaHBx<m

 

 

mpzomw PseseHSMoem osp one po% Egon :ewa omH .em
And Cav mHo>m4 bsflpmem ososumsZH soawosesoo pfl< exp %0 otssm tee .ssflcea .ssea

{tr e .l

103

 

ooumm osumm muses

 

Ham me mm .mp me sfi mcaozmopge COHPDooop mooonm
mmnoe mwusm steam
30.5: Ho.mm Emma

-Aseaav shoe omH
.op mo CH Amm ooom use .oooa .oomv meaosmopcP esmpm>s meow ease

 

apomw MEIHw4uo mpomm H<H8m¢m

 

 

.mesopm Pemseflzpoom 038 one he
now mUHOSmoseB coappooox goooem mo new A.Nm ooom Use
.HV

come now see whoa one
. mwaosmepme mumsm>< ecosnessa mo mehssm use meme;

..Nm oooa ..Nm com m

APPENDIX C

CONPRESSOR INPUT-OUTPUT FUNCTIONS

1011'r

Input-cutout Functions (in dB) of the Thr;e Contressor
Channels when the CNC Words were Comnressed under Cnc—to—
one, Two-to-one, and Three—to-one Ratios.

 

 

Cne—to-one Comnression Ratio

 

Freeuencies (Hz)
level (in d?) 700 1500 3000

 

o o 0 0
~10 -9.6 -9.u -9.0
—20 -19.M - 9.3 -18.7
-30 —2s.6 —¢3.7 -2s.3

 

 

O O

\nfio
{ECU
l
I HI
.‘OknO
I 0
WC

I-"I
JL.“
0
O'
I
._I
N.
.

 

Heression Ratio

(3
O
”3

Three-to—one

 

-20 ..
-3o _;

I
H
O
I
Q "\J'. {\J O
I
(\3
K
I
\J) O

0

kn O
l

O k

I

\f‘ (“‘\
.é‘.

 

db

0"

E.

4:5,.“

I"

 

 

4H «aw «HHH H_H mHH QHHH H«m mHHH UHH H.m 4H ma
<H OHHH mHH H.N mHHH dHH H.H <HHH OHH Hum «H NH
<H mHHH QHH H.H mHH QHHH H.m <HH mHHH H.m 4H ma
<H UHHH DHH H.m <HHH <HH H.H mHHH mHH H.N <H ma
<H <HH QHHH H.N «HHH UHH anm mHH mHHH H.H <H i
¢H mHH «HHH Ham QHH OHHH H.w mHHH <HH H.H <H ma
4H <HH <HHH H.H mHH QHHH H.m mHHH UHH H.m 4H NH
eH <HH mHHH H.m <HHH mHH H.H QHH UHHH H.m 4H Ha
<H mHH OHHH Hua mHHH UHH H.m <HHH <HH H.m mH OH
sH QHH mHHH H.m UHHH <HH H.H mHH <HHH H.m <H m
<H OHHH ¢HH H.m mHH mHHH H.m UHH <HHH H.H <H w
«H UHHH «HH Hum mHH <HHH H.m mHHH QHH H.H sH m
«H dHH DHHH H.H UHH mHHH H.m <HHH mHH H.m <H w
dH OHHH mHH H.N mHHH <HH H.H fiHHH OHH H.m 4H m
<H <HHH UHH H.H mHH OHHH H.m <HH mHHH H.N 4H 3
<H OHHH QHH H.m fiHHH <HH H.H mHHH mHH H.m 4H m
<H <HH UHHH H.m «HHH OHH H.m mHH mHHH H.H 4H N
dH mHH <HHH H.m UHH OHHH Hum mHHH mHH H.H «H H
same ewes
lemom memea oases memflg oeedg mpmflg oeesm teem meommpsm

 

 

 

.memfla ewes one

PCmEPHSMooM ose esp we now
.msoflpflesoo Scam

m CH
mops

m
E0

90
o

.w950pm

omQSm seepguflm one hoe eweenpmoe exp new

we-

.emmenopp ens sow nacho :oflemesemepm

TECTS PCP. CNC WCRDS

A1- 7‘L’7‘ .4DIX E
LTIEYI‘ erCr1AllC‘ InELIh
ND DIS CRIBE IIATICN SCCR:E S CF
U3

IN
7'1

In

r‘CR IIE

ARY RESECN
E THIRTY- SIX
TIIIJQE 11:11 wIICS

SES,

106

 

 

N am mm RH 9H sH am mm mm mH om om em om H a m m
em Hm ma mm Hm 0 RH sH mH om om OH mm mm H a we BH
3 mm mm em em om am mm on mH om mH mm a: x s Hm mH
mm mm Hm mm mm em mm mm om om mH mH em a. H m on mH
a: om om me we es we we mm om mm mH m: mm H a Hm e-
mm mm mm om wH mH Hm sH mm om om on we cs H a we mH
NH HH RH mH sH 0H eH mH mm om mm mm my cm H a He NH
an mm N am om mm mm om mm om om mH mm mm m m mm HH
we we we a: om we we HH om 0H mH 0H em mm H a mm OH
on mm mm Hm om mm am am om mH OH OH em an H a e: 0
mm mm om sH @H mH m nH om mH mH mH ow om x a mm m
on mm om mm on mm mm mm OH on o m mm mm H a me a
oe mm mm mm em mm mm mm om mm mm om mm mm H a mm m
A: m: 03 w 3 N: m3 Hi OH OH m m mm 0m H A mm m
an m: H: o: H: mm mm a mm om om om ow mm x a Hm a
H: mm H: on on on em Hm mm om mm om mm as m s om n
me He me He 0: 9 me as mH om mm mH om om e m mm m
mm H: Hm Hm mm mm on om mm mm mm om me co m s we H

PmQE Pmmrw om OH Adm.

m H N H m
H.m H.m H.H
mpoom Bax COHP CHEHpomHQ zooonm

 

H .
H upmoL -mCH aocm ooom QQOH com «em eam :emme emm sea mm
H v seem mHme sea

 

 

mOHme COHmmospsoo moses one

he a memo: 030 one soy mpoompsm HmesesHpe-N sHmlaepHea
one %o mopoom COHPssHEHsom

Us
HQ gem .momsoemom zpwcHEHHmsH .QOHPmEhomsH msHaeH92oeH

Hpumw BMMAB axomm H<HEJ<L

(

 

 

 

 

 

lo?

 

 

 

O: we as me me O; O: He mN ON ON ON mm O: H a me OH
mm mm me N: an N 1 mm mN mN mH ON we N: H 3 ON NH
Os me me we as He me as ON mN ON mN we Om N H mm OH
mm it im at mm 0m mm mm mH mH mH om we m: x H mm mH
H mm mm mm eN NN Hm ON OH om mH -N we es x m mm OH
He me as me mm in H: Os ON ON ON ON me Om H m NO NH
as he as O; mm ON mm ON ON ON ON ON Nm ON a H HO NH
me me me 0: mm H: O- Nm ON ON OH OH Os Om H a He HH
Ne as mm o: Oe Om Om mm mm OH mH ON O: as x m O, OH
Ne as me we Ne O: we as mN OH OH OH N; Om H H mm O

mm NO ON NO ON ON N NN mN mH mN mN Om Nm H H mm O

as Om O: we Hm Om He NO ON ON mH mH Ne Nm m a m- N

me Os Ne Ne HO Oe H: m- ON mN ON mH .m em H a OO O

me me He Ne NO NO ON ON mH ON OH m me me H m NO m

ON NN ON ON NN ON NN NH ON ON mN m- we Om H n NN e

Hm eN mm NN Hm NN mm Hm Om mH mN mN Om OO m a N: m

me me me O: N3 3: m: ON ON OH mH mN we we H a Rm N

me me HO HO eN ON NO H: ON ON ON ON Om we H a a: H

Pmoa ewes mmHoz New

N H N H N H :emom -msH seem OrON OOOH Opm 43H gem seems New sea mm
e.m H.N H. NN.NHO when mHm< rum

opoom 3mm COHHNCHEHmomHQ eoooem

 

 

 

 

mOHHNm COHmmopesoo omega map News; mono; ozo esp Noe meoomesm HOPsoEHNoQNm melzpmHza
mew e0 mmpoom COHPNCHEHHomHQ use .mmeQCmom RNNCHEHHopm .COHPNEpoecH msHaersmOH

amaze BZmHHHDmomx MBmHmauo