A STUDY OF ENERGY VARIATIONS OF
TWELVE VOWEL SOUNDS USING THREE
METHODS OF PICKUP WITHIN
THREE BANDWIDTHS

Thesis for Hm Degree of M. A.
MICHIGAN STATE UNIVERSITY

Kenna Collins

1963

 

LIBR A R Y
Michigan State

K.
L, University (3

V‘WJ‘FM

 

 

 

 

' 0 I . ‘ ' ' .‘\.,.o r

jimhw ‘2
W n

 

 

ABSTRACT

A STUDY OF ENERGY VARIATIONS OF TWELVE VOWEL
SOUNDS USING THREE METHODS OF PICKUP
WITHIN THREE BANDWIDTHS

by Glenna Collins

The purpose of this study is to explore the energy
variations at each of three bandwidths for twelve vowels
of the English language simultaneously recorded at the lips
and forehead of six different speakers using three types
of pickup devices.

The subjects for this study were six air cadets at
the Naval Air Station, Pensacola, Florida. They were asked
to intone the twelve vowel sounds [£,I,e,£,&,a,) , O,
'U', L(,/\, and 3‘ ] which were then simultaneously recorded
from pickup points at the forehead by a bone oscillator and
condenser microphone, and by a condenser microphone at the
lips.

The findings of this study indicate a greater mean
vowel amplitude for air conduction. They also indicate a sig-
nificant difference in the mean amount of energy between
vowels for bandwidths I, II, and III (100—199 cps, 200—499
cps, 500—999 cps), between pickup methods for bandwidths I,
II, and III, and a significant interaction between vowels
and points of pickup for bandwidths I and III at the 5 percent
level of confidence. The significant difference between any

two vowel sounds may be due to some interaction taking place

Glenna Collins

between the two vowels in question and the points of pickup
or may be a function of the pickup points. The findings also
indicate the relationship between any two points of pickup
and the amount of energy for any given vowel.

The conclusions which were drawn from this study
suggest that air conduction is a better method of pickup
than bone or tissue conduction as far as vowel amplitude is
concerned. The front vowels tend to be alike in mean energy
in all of the bandwidths. Also, most of the significant
relationships between any two points of pickup and the

amount of energy for any given vowel are found in bandwidth I.

A STUDY OF ENERGY VARIATIONS OF TWELVE VOWEL
SOUNDS USING THREE METHODS OF PICKUP
WITHIN THREE BANDWIDTHS

By

Glenna Collins

Submitted to
Michigan State University
in partial fulfillment of the requirements
' for the degree of

MASTER OF ARTS
College of Communication Arts, Department of Speech

1963

 

TABLE OF CONTENTS
Page
LIST OF TABLES. . . . .‘. . . . . . . . . . . . . . . . iii

LISTS OF FIGURES. . . . . . . . . . . . . . . . . . . . iv

Chapter
I. STATEMENT OF THE PROBLEM. . . . . . . . . . . 1

Introduction

Statement of Problem and Purpose
. of Study

Hypotheses

Importance of Study

Definition of Terms

II. REVIEW OF THE LITERATURE . . . . . . . . . . 5

III. SUBJECTS, EQUIPMENT, MATERIALS, AND
PROCEDURES . . . . . . . . . . . . . . 12

Introduction
Part I
Part II
IV. RESULTS AND DISCUSSION . . . . . . . . . . . 17
Introduction
Results
Discussion
'V. SUMMARY AND CONCLUSIONS . . . . . . . . . . 41
Summary

Conclusions
Implications for Future Research

BIBLIOGRAPHY O O O O 0 O O O O O O O O O O O O O O I O O 51

ii

Table

II.

III.

VI.

VII 0

VIII.

LIST OF TABLES

Mean Amplitude of Twelve Vowels in Each of Three
Bandwidths for Three Pickup Points Within Each
Bandwidth. . . . . . . . . .

Analysis of Variance of the Mean Amounts of Energy
in Bandwidth I of the Twelve Vowels Under Three
Methods of Pickup. . . . . . . . . . . .

Analysis of Variance of the Mean Amounts of Energy
in Bandwidth II of the Twelve Vowels Under Three
Methods of Pickup. . . . . . . . . . . . . . .

Analysis of Variance of the Mean Amounts of Energy
in Bandwidth III of the Twelve Vowels Under
Three Methods of Pickup. . . . . . . . . . .

Table of Differences for Pairs of Vowels in
BandWidth I 0 O O O O o O O O 0 O O O O O 0

Table of Differences for Pairs of Vowels in
Bandwidth II . . . . . . . . . . . . . . . . . .

Table of Differences for Pairs of Vowels in
Bandwidth III.

Presentation of Correlation Coefficients Between

all Combinations of Two Points of Pickup for
all Twelve Vowels in the Three Bandwidths. . . .

iii

Page

19

. 21

.23

24

26
29

32

37

LIST OF FIGURES

Figure Page

1. A Schematic Representation of the Arrangement
of Recording Equipment. . . . . . . . . . . . . . 14

2. A Schematic Representation of the Equipment
Employed. . . . . . . . . . . . . . . . . 15

iv

CHAPTER I

STATEMENT OF THE PROBLEM

Introduction

 

Past studies indicate that intelligible speech can be
recorded from various locations on the Speakefis body other than
the lips. Several studies have been done either directly or
indirectly on tissue transmitted speech, and some have appeared
to have merit as far as intelligibility of speech and listener
preference are concerned.

The problem under investigation in this study is to
compare the energy variations at each of three bandwidths for
twelve vowels of the English language simultaneously recorded at
the lips and forehead of six different speakers using three
types of pickup divices. Knowledge in this area should provide
additional insight as to the possibility of using body trans-
mitted speech in place of Speech picked up at the lips of the

Speaker.

Statement of Problem and Purpose of Study
The problem from which this study arose is that of deter—
mining the intelligibility of tissue transmitted speech with
different types of pickup equipment. The purpose of this
study is to analyze the results obtained from exploring system-
atically the energy variations of twelve vowel sounds picked

l

up at the forehead and the lips using a condensor microj

phone and bone oscillator at the forehead and a control

micrOphone for the mouth emitted signal pickup.

From this analysis it is hoped that answers to the

following questions can in part be obtained:

1.

Does the mean amount of energy differ from vowel
to vowel within each of the three bandwidths?
Does the mean amount of energy differ as a
function of pickup methods within each of the
three bandwidths?

Is there an interaction between vowels and points
of pickup within each of the three bandwidths?
What is the relationship between any two points
of pickup and the amount of energy for each

vowel?

Hypotheses

 

These questions can be formulated into the following

null hypotheses:

1.

There is no significant difference in the mean
amount of energy between vowels within each of

the three bandwidths.

There is no significant difference in the mean
amount of energy between pickup methods within
each of the three bandwidths.

There is no significant interaction between vowels
and points of pickup within each of the three

bandwidths.

A. There is no significant correlation between the amount
of energy for any given vowel and any two points of

pickup.

ImpOrtance of Study

Several studies relating directly or indirectly to tissue
transmitted speech have been reported. Numerous anatomical sites
have appeared to have merit as far as intelligibility of Speech
and listener preference are concerned.

This study was designed to explore systematically the
variations in energy signals of Speech picked up at the lips and
the forehead. It will prOvide additional information concern-
ing the energy of vowels for different bandwidths and give
insight as to the possibility of using the forehead as a point
of pickup of Speech.

This type of information could be of value to those work-

ing<n18pecial projects concerned with transmission of Speech

signals.

Definition of Terms

Bandwidth

 

A frequency band expressing in cycles per second the
range of frequencies from lOO—9,999 cps within which six di-
visions were made in the original study as follows: 100-199 cps;
200-499 ops; 500-999 ops; 1000-1999 Cps; 2000-4999 CpS; 5000-

9999 CpS.

Decibel
A relative unit of power using .0002 dyne per cm2 as a
reference level which increases as a logarithm of the ratio of

the greater intensity to the lesser intensity.l

Bone Conduction Pickup

 

A method of picking up sound waves that are conducted
via bone using a bone oscillator on the frontal bone of the

Speaker's skull, one inch above the nasalis.

Tissue Conduction Pickup

 

A method of picking up sound waves that are conducted
via tissue using a condensor micrOphone (Altec 21 D) at the

forehead of the Speaker.

Air Conduction Pickup

 

A method of picking up sound waves that are conducted via
air from the lips using a micrOphone (Altec 21 D) positioned

approximately eighteen inches in front of the speaker's mouth.

 

1Giles w. Gray and Claude M. Wise, The Bases of Speech
(New ”York: Harper & Brothers, 1959), pp. 106-107.

 

CHAPTER II

REVIEW OF THE LITERATURE

During the past thirty-six years some interest has been
manifested by several investigators in the recording of vocal
and Speech signals from various locations on the body of the
Speaker. Some investigators have designed experiments in order
to measure the intelligibility of the tissue transmitted signals,
whereas others have merely mentioned their observations of Such
signals while in the process of exploring something else.

As far back as 1926, Robert West reported a study of the
nature of vowel sounds utilizing a stethoscope at various anatom—
ical locations to listen to sounds as they were produced;2

In 1927 an article in the Quarterly Journal of Speech

Education written by Clarence Simon and Franklin Keller

 

described in detail how seven areas of the body were examined
and how the vibratiOns were photographically recorded as the
subject intoned the vowe1[O].

Voice and chest waves were recorded simultaneously by
meenqs of phonosc0pes. Contact with the body was made with a

carfloon micrOphone, with the diaphragm removed. The protruding

 

2Robert West, "The Nature of Vocal Sounds," Quarterly
Journual of Speech Education, 12 (1926), pp. 224-293.

5

pin picked up vibrations within a restricted area of three
inches around the point of contact. In all subjects, vibrations
picked up from the right wing of the cartilage followed most
closely the cord tone.3
In 1932 an investigation of chest resonance, as reported
by Claude Wise, showed that the conductile effeciency of human
tissue varied and decreased in the following sequence: (1)
body tissue, (2) tendinous tissue, (3) tense muscle tissue,
(4) relaxed muscle tissue, and (5) soft nonmuscular tissuef‘L
In the early thirties, the Signal Corps Acoustical Lab-
oratory, Fort Monmouth, New Jersey, was interested in developing
a micrOphone that could be used in noisy environments and could
be worn in such a manner to leave both hands free. The late
George Graham, a section chief at that time suggested that
a sound powered receiver unit might be employed as a receiver
and as a microphone driven by the sound transmitted via the
Eustachian tube. Information concerning this research project

was obtained by Oyer in communications with those associated

with the research.5

 

3Clarence Simon and Franklin Keller, "An Approach to the
Problem of Chest Resonance, "Quarterly Journal of Speech Education,

13 (1927), pp. 432-439.

“Claude M. Wise, "Chest Resonance," Quarterly Journal of

Speech, 18 (1932), pp. 446—452.

5Herbert J. Oyer, "Relative Intelligibility of Speech
Recorded Simultaneously at the Ear and Mouth," The Journal of
the Acoustical Society of America, 27 (November, 1955), pp.
1208-1209. I

 

Harry W. Parmer (in early thirties), Chief, Advanced-
Systems Engineering Group, recalled (in 1955) that a bone
conduction receiver was modified to provide a small diaphragm
for driving the receiver unit. When prOperly placed in close
contact with the skin, the diaphragm was fairly effective in
excluding external noise. Highest output was obtained when the
unit was employed as a throat micrOphone, but the best Speech
quality was derived when the unit was located on the cheek
close to the tOp of the ear or directly in front of the lower
part of the ear. The possibility of utilizing cheek units,
however, was abandoned due to mounting difficulties.

Albert E. Woodruff (in the early thirties, a member of
the Research Department, Automatic Electric Company) recalled
that the Company was requested by the Signal Corps to design
an eSpecially sensitive receiver for use as an ear micrOphone.
His observations led him to believe that an ear microphone was
not successful if it were pressed tightly against the ear. This
led him to believe that successful use of the~ear microphone was
largely dependent uponair-bournes;eech signals. The use of
effective ear cushions and other means of decreasing air-bourne
signals seemed to confirm this notion, thus leading to the
conclusion that no more speech energy could be picked up from

the ear canal that from the skull elements surrounding it.7

6Ibid.

71b1d.

In 1949 James Mullendore reported on a study of the relative
amplitudes of sound vibration at various body locations during
sustained productions of vowel sounds. This study showed the
composite rank of intensity at ten micrOphone positions to be:

(1) thyroid cartilage, (2) mandible, (3) nose, (A) top of head,
(5) clavicle, (6) vertebra, (7) sternum (superior end), (8)
sternum (inferior end), (9) mastoid, (10) fifth rib.8

In 1951 von Békésy and Rosenblith reported in the 522E?
book of Experimental Psychology that vibrations of the surface
of the skin that are recorded by means of a pickup at various
body locations while the Subject sings a vowel that has a
displacement amplitude in the vicinity of the ear is only one-
twentieth of the amplitude recorded near the vocal cords. The
attenuation between the oral cavity and the ear canal was re—
ported to be forty to fifty decibels.9

Hirsh and Benson in 1952 stated in a WADC Technical Re-
port that the ear canal does provide a source of sound pressure
which can be utilized for delivery of speech from talker to
listener in a communications system. Sounds from this source
appear to have quality as good as sounds picked up at the lips

of the talker.10

 

8James M. Mullendore, "Relative Amplitudes of Sound Vi—
brations at Various Body Locations," Speech Monographs, 16
(1949) pp. 163-177.

98. S. Stevens (ed.), Handbook of Experimental Psychology
(New York: John Wiley & Sons Inc., 1951), pp. 111.

10I. J. Hirsh and R. W. Benson, "Wright Air DevelOpment
Ckniter, WADC Technical Report No. 52, 175,' (May, 1952), pp. 20—21.

In 1955 0yer carried out a study to determine the rela—
tive intelligibility of simultaneously recorded speech signals
picked up at the lips and left ears of speakers. This study,
as recorded by 0yer, showed that as listening conditions
became more difficult the signal picked up at the ear canal
was more intelligible than simultaneous recordings at the
lips.11

The need for improved micrOphones and receivers in
aviation was pointed out in the 1956 study reported by Moser
and Dreher. They stated that both transmission and reception
of signals would be improved by lightening the weight and
simplifying the radio equipment. Hygiene and sanitation
would be improved by equipment inexpensive enough to allow
personal ownership of components. The ear and bone units
used as microphones in this experiment proved to be equal
in performance to those presently used.12

In the course of ear-signal investigation (from 1955-
1957) at Ohio State University Psycholinguistics Laboratory
a report of relative intensities of sounds at various
anatomical locations of the head and neck during phonation

of the vowels was made by Moser and Dyer. Results Showed

the signals that were most intense were picked up immediately

v—

ll0yer, op. cit.

12Henry M. Moser and John J. Dreher, "Operational Tests
of Miniature Microphones and Receiver" (Technical Report No.
36", AFCRC TN 56-57, (Oct., 1956), pp. 1-3.

10

below the superior thyroid notch of the laryngeal prominence
and least intense signals were picked up from the squama of
the temporal bone one inch above the external auditory
meatus. The forehead signal was weak but quite clear.13

In 1958 0yer, Moser, and Wolfe did a further analysis
of the 1955 0yer study. It was the purpose of this study
to analyze differences in listener's reSponses to Speech
signals picked up in the ear and in front of the lips. It
was found that more confusion occurred for ear-recorded
than for lip—recorded Speech Signals. AS the listening
conditions became more adverse, the listeners tended to omit
more of the words recorded at the lips but continued to
make attempts to reSpond to those recorded at the ear.
Words containing the[¥{)sound were significantly more intel-
1igible when the stimulus was recorded at the lips whereas

the Opposite was true for“: , 6’6 and Iivowels. There

:

was no apparent relationship between consonant structure and

intelligibility of the test words. It appears from this

study that a trained listening panel takes more cues from

the vowel and dipthong components than from the consonants.lu
A preliminary study reported by Snidecor, Rehman, and

Washburn in 1959 investigated the relative power of the vowels

 

13Henry M. Moser and Herbert J. 0yer, "Relative Inten-
sities of Sounds at Various Anatomical Locations of the Head
and Neck during Phonation of the Vowels," The Journal of the
Acoustical Society of America, 30 (April,l958), pp. 175-177.

1”Herbert J. 0yer, Henry M. Moser, and Susan M. Wolfe,
”Relationship of Phonetic Structure to the Intelligibility of

ll

51:, E, 9 ,UPI‘li-i the relative quality preference for a
standard sample of continuous Speech. The locations of
forehead, mastoid process, larynx, mandibular angle, ear
canal, and nose give promise of being suitable pickup
positions during military duties requiring that no lip or

free—field micrOphone be used.15

 

Words Simultaneously Recorded at Ear and Li s," Journal of

Speech and Hearing Research, 3 (March, 1960 pp. 44-50.

 

 

15John C. Snidecor, Irving Rehman, and David D. Wash-
burn, "Speech Pickup by Contact MicrOphone at Head and Neck
Positions," Journal of Speech and Hearinngesearch, 2

(Sept., 1959)-

 

CHAPTER III

SUBJECTS, EQUIPMENT, MATERIALS, AND PROCEDURES

IntroductiOn

 

The material in this chapter is divided into two parts.
In Part I is presented an explanation Of the experimental arrange-
ments pertaining to the simultaneous recording of the vowel
sounds.16 In Part II is presented an explanation of the

experimental arrangements involved in the physical analyses of

the acoustic properties of the vowel sounds.

Part I
I. Subjects
The subjects participating in the original study were
air cadets at the Naval Air Station, Pensacola, Florida.
There were six in all.

II. Equipment

 

The equipment employed in the original study were as
follows:
1. Tape Recorder

2. Two Condenser MicrOphones (Altec 21B)

 

16Information derived from personal interviews with H. J.
0yer. July, 1962; January, 1963.

12

l3

3. Bone Oscillator (Dyna—Jet)

A. Helmet (Naval Flight Type)

III. Material

IV.

The material utilized in this study consisted of twelve
vowels. They were as follows: [L ,I, 9,8,39, Q7 , O,‘U‘,

u,/\,3‘i.

Procedure

 

The procedures employed in the original study were as
follows:

Recording

 

Each of the six subjects was requested to sit in a chair
that had attached to it a head stabilizer. A helmet that
contained one condenser micrOphone and a bone oscillator
was then placed on the head of the subject. A condenser
microphone was placed eighteen inches from the lips of the
speaker and was mounted on a microphone stand. The subject
was asked to intone the twelve vowel sounds listed above.
With this arrangement it was possible to effect a Simul-
taneous recording of vowel sounds picked up at the forehead
by the bone oscillator and the condenser microphone at the
same time the vowel sounds were being picked up at the lips.

Figure 1 presents a schematic representation of the
arrangement of recording equipment.

Part II
The recorded vowel sounds were then fed to an analyzer for

anualysis of acoustic components. The type of equipment

.hoxmoam some now hapnwaam Umpmznnm one: mHo>oq

.D> OHON coospoha Hoccmno some page on H noxmoam so

you mm: Ho>oq.mo::ow Hozo> mo wadopooom msoozmuasefim :H
UONHHHDS pcoaomcmhhm acoEQHSUO mo coapmpcomomaom OHmeonomln.H madman

l4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

coaeaaae< Hmmmum
Hoccmnonm Koofi<
- - ADEmmHBV
_ OCOQQOLOAE
_@ u MW pomcopcoo
- poemaaaomo
_ ,l\\HI .M\
HMNfi! //L nu ocom
A V 73.5
BLT /_ Rm ozonoonoaz
noxmoam poumscopp< noncoocoo
Honcmnonm Hoccmnonm Eoom pmoe

 

 

 

 

 

l5

.Hoccmno ANOCHH CH HHH Log 0: pm

wow poohooom Ho>oq .HHH pom D>m| pm wow Ho>oq @009

came .Hozo> on» mo mfimhamcm pawn o>mpoo phanpuoco
on» EH GOSOHQEO acoEQHSUO mo cofipmpcomonaon OHmeonomnu.m opswfim

mopaooom popoaoupooam

 

 

 

 

 

 

 

soon spa:
momm
xoae<

Monaco:

 

 

 

 

 

Ho>oq md
sommxum monMIm
r nnnnnnnnnn u

owmeHq

 

 

 

 

somnscfim
Hoscmco-m
xoae<

 

 

 

 

l6

employed permitted a one third octave band analysis. Figure
2 presents a schematic representation of the equipment
employed.

The paper tapes from the level recorder were then
analyzed. These tapes were divided vertically into six
bandwidths expressed in cycles per second from loo-9,999
and horizontally into decibels. A tape was made as each
speaker intoned a vowel sound covering the three methods
of pickup. Consequently there were thirty—six tapes for
each of the six speakers or two hundred-sixteen paper
tapes to be analyzed in all.

The mean energy was tabulated for each bandwidth by
finding the average of the three highest peaks of the
tracings within each bandwidth. These raw scores were then
charted by speakers for each method of pickup over all
twelve vowels. Since it was evident at this point that
there were mostly zero scores for the last three band-
widths, only three bandwidths were included in this study
with 100-199 cycles per second being referred to as band-
width 1, 200—499 cycles per second being referred to as
bandwidth II, and 500—999 cycles per second being referred

to as bandwidth III.

CHAPTER IV

RESULTS AND DISCUSSION

Introduction

 

The data necessary to begin this study were obtained by
analyzing the paper tapes taken from the level recorder. The
mean amplitude in decibels was found for each bandwidth and these
data were charted according to vowels, Speakers, and points of
pickup. A separate analysis of variance was done for each of
the three bandwidths utilized in this study in order to deter-
mine whether or not a significant difference was evident between
vowels and between points of pickup and whether or not there
was a significant interaction between vowels and points of pick-
up at the 5 percent level of significance.

A test of individual comparisons was applied to all pos-
sible pairs of means to determine which means were significantly
different in each of the three bandwidths over all three methods
of pickup.

Correlation coefficients were worked out between each
possible pair of pickup points for each of the twelve vowel
sounds to establish the relationship between the points of

pickup and the amount Of energy for any given vowel.

l7

18

Bandwidth

 

Upon inspection Of the raw scores it was evident that
very little acoustical energy was contained in the last three of
the six bandwidths. Because the mean-amplitudes for these last
three bandwidths were close to zero, they-were omitted for not
contributing anything to the analysis. For this reason, only
three bandwidths were utilized in this study. These bandwidths
expressed in cycles per second were 100-199, 200-499, and 500—
999 and shall be referred to as bandwidths I, II, and III

reSpectively.

Raw Scores

 

Raw scores were Obtained for each of the bandwidths by
computing the mean amplitude or energy in decibels using the
three highest peaks within each of the bandwidths. The raw
scores for each of the three bandwidths used were placed
according to vowels and points of pickup, with each of the twelve
vowels having six values, one forieach Speaker for each Of the
three points of pickup. Thesé/raw scores are presented in

Appendix I.

Results

 

From Table I it is evident that the mean vowel ampli-
tude is greater for air conduction than for bone or tissue
conduction in all three bandwidths. One might expect air con-
duction to haVe the highest mean vowel amplitude since air is

known.to be a good conductor of sound waves. Then one would

II

III

AC
BC
TC

AC

TC

AC

TC

. MEAN AMPLITUDE (db re:

19

'TAEHJE I

0.0002 dyne/cm2) 0F TWELVE

VOWELS IN EACH OF THREE BANDWIDTHS FOR THREE
- PICKUP POINTS WITHIN EACH BANDWIDTH

 

33.2

ES

82

32

6L

9

£1.

 

32.63l.1

29.3

0

I.

3

29.228.8

31.2

.;1£_
32.8

.144

/\

3F

Mean
Vowel
Ampli-
tude

 

34.0

31.1

31.1

31.14

 

21.4

l7.718.2

l4.212.8

13.814.7

18.020.3

21.1

15.5

17.5

17.10

 

28.8

23. 24.5

20.1

9.0

19.421.3

25.626.8

29.1

-2.1

25.0

23.80

 

 

32.2

37. 36.8

32.230.2

29.231.3

38.139.5

37.2

32.2

37.

34.45

 

32.6

32.633.7

26.623.4

24.225.3

32.934.3

34.5

27.0

31.8

29.91

 

30.5

29.930.4

23.020.8

21.522.0

29.382.3

32.4

2.6

29.7

27.03

 

 

11.0

23.722.5

33.136.0

40.441.3

33.329.6

23.2

38.8

30.5

30.28

 

10.8

l9.él9.0

26.127.5

31.333.2

27.924.2

19.4

29.2

24.0

24.35

 

4.6

 

6.1 5.4

 

 

7.9

8.2

 

 

11.013.3

 

 

12.5

9.4

 

7.4

 

10.8‘

 

 

9.2

 

8.82

 

 

 

Ac
Bc
TC

air conduction

bone conduction
tissue conduction

 

2O

probably expect bone conduction to have the next highestwmean~
vowel amplitude and tissue conduction to have the~lowest~since
bone or any solid substance with a high-density is a better
conductor of sound waves than tissue Of unequal density and
resiliency. However, from Table I it can be seen that band-
width I did not follow this pattern in this study. Further
research is needed in this area to Obtain any definite conclusions.

A separate analysis of variance was done for each of the
three bandwidths. The differences in column means, (vowels)
and the differences in row means, (points of pickup) were in-
vestigated utilizing an analysis of variance design described
by Dixon and Massey as apprOpriate with two variables Of
classification and repeated measurements.17

The analysis in Table II reveals that there is significant
difference in the mean amount of energy between vowels at the
5 percent level of confidence. There is also a significant
difference in the mean amount of energy between pickup methods
at the 5 percent level of confidence and a highly significant
interaction between the vowels and points of pickup.

Therefore, in view of this information, the first hypoth-
esis which states that there is no significant difference in
the mean amount of energy between vowels for bandwidth I would
be rejected. The second hypothesis which states that there is

no significant difference in the mean amount of energy between

 

l7Wilfrid J. Dixon and Frank J. Massey Jr., Introduction
to Statistical Analysis (New York: McGraw-Hill Book COmpany

InC., 1957), pp. 163-168.

TABLE II

21

ANALYSIS OF VARIANCE OF THE MEAN AMOUNTS OF

ENERGY IN BANDWIDTH I OF THE TWELVE VOWELS

UNDER THREE METHODS OF PICKUP

 

 

 

 

 

 

 

 

 

Source Sun10f Squares df ‘Mean Square F
Row Means
or 7,035.24 2 3,517.62 2,883.3*
Pickup Points
Column Means
or .1,4l2.37 11 128.40 105.2**
Vowels
Interaction 4,253.74 22 193.35 158.5x
Subtotal 12,701.35 35
Within Groups 220.45 180 1.22
Total 12,921.80 215

 

 

 

 

 

*With df of 2 and 180 and F of 3.00 is required for
level of confidence.

significance at the 5

**With df of 11 and 180
significance at the 5

xWith df of 22 and 180
significance at the 5

percent

percent

an F of
percent

an F of 1.79 is required for

level of confidence.

1.54 is required for
level of confidence.

 

pickup methods for bandwidth 1 would also be rejected, and the
third hypothesis which states that there is no significant
interaction between the vowels and pickup points for bandwidth
I would be rejected.

The analysis shown in Table III reveals that at the 5
percent level of confidence there is a significant difference
in the mean amount of energy between vowels. There is also a
Significant difference in the mean amount of energy between
pickup methods. However, the data shows a non-significant
interaction between the vowels and points of pickup.

The presentation of data for bandwidth II reveals that
the first and second hypotheses which state that there is no
significant difference in mean amount of energy between vowels
and between pickup methods would be rejected. Results indicate
a failure to reject the third hypothesis because there is no
significant interaction between the vowels and points of
pickup for bandwidth II.

Table IV reveals that there is a significant difference
in the mean amount of energy between vowels and between points
of pickup at the 5 percent level of confidence. There is also
a significant interaction between the vowels and points of
pickup at the 5 percent level of confidence for bandwidth III.

In view of this information, the first, second, and
third hypotheses would be rejected for bandwidth III.

The F—Test in this analysis is regarded as an over—
all test of significance determining whether the amount of
energy for any one or more of the twelve vowels is significantly

differerrt from the amount of energy for any of the other twelve

 

TABLE III

.ANALYSIS OF VARIANCE OF THE MEAN AMOUNTS OF
ENERGY IN BANDWIDTH II OF THE TWELVE VOWELS
UNDER THREE METHODS OF PICKUP

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(Source Sum of Squares df Mean Square F
Row Means
or 2,020.22 2 1,010.11 76.9*
.Pickup Points
Column Means
or 3,189.20 11 289.9 22.1**
Vowels .
_Interaction 207.17 22 9.4 .72x
Subtotal 5,416.59 35
Within Groups 2,363.89 180 13.13
.Total_ 7,780.48 215
*An F of 3.00 is required for significance at the 5
percent level of confidence with df of-2wand 180.

**An F of 1.79 is required for significance
level of confidence with df of 11

percent

XAn F of 1.54 is required for significance
level of confidence with df of 22

percent

at the 5
and 180.

at the 5
and 180.

 

24

TABLE IV

ANALYSIS OF VARIANCE OF THE MEAN AMOUNTS OF
ENERGY IN BANDWIDTH III OF THE TWELVE
, VOWELS UNDER THREE METHODS OF PICKUP

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Source Sum of Squares vdf :Mean Square ‘ F
Row Means
on. 17,140.25 2 8,570.12 310.06*
Pickup .
Points f
Column Means 1 '
or 11,011.04 11 1,001.00 36.21*
Vowels
Interaction 2,074.39 .22 94.29 3.42x
Subtotal 30,255.68 .35
Within
Groups 4,975.25 180 27.64
Total 35,200.93 215
*An F of 3.00 is required for significance at the 5
percent level of confidence with df of 2 and 180.
**An F of 1.79 is required for significance at the 5
percent level of confidence with df of 11 and 180.
xAn F of 1.54 is required for significance at the 5
percent level of confidence with df of 22 and 180.

25

vowels. It does not necessarily follow-that the meannamount
of energy for each of the twelve vowels differs from all other
vowels. Therefore, a test of individualwcomparisons was applied
to all possible pairs of means to determine which means were
significantly different in each of the three bandwidths over
all three methods of pickup.

The critical difference was computed corresponding to the
5 percent level. The mean amount of energy for each vowel was
compared with each other vowel in bandwidths I, II, and III
over all three methods of pickup with the results presented in
Tables V, VI, and VIIreSpectively.18

 

18
E. F. Lindquist, Design and Analysis of Experiments in
Ppychology and Education (Boston: Houghton Mifflin Company,

1953): Pp. 90-90.

5% (n (D 14 E‘-

AC

co (D *4 ?~.“%.2> S2

53

BC

TC

A1.)»:

TABLE OF DIFFERENCES FOR PAIRS OF VOWELS

.6 2.1* 3.9*

1.5* 3.3*
1.8*

—-v

TABLE V

AIR CONDUCTION

W...“ - -. -o.

4.0* 4.4%

3.4* 3.8*

1.9* 2.3*
.1 .5
.1 .5
.4

2.0*
1.3*
.2
1.9*
1.9*
2.0*

2.5*

IN BANDWIDTH I

.8 2.1*
1.4* 1.5*
2.9* 0
4.7* 1.8*
4.7* 1.8*
4.8* 1.9*
5.2* 2.3*
2.8* .1

1.2* 1.7*
2.9*

With a critical difference of 1.25 at the 5 percent
level of confidence, the values for the pairs of
vowels marked with an asterisk show a significant

difference.

 

AC

BC

TC

.4 11.8* 15.5%
:r 11.1% 14.8*

e
El
39

7

.fl

00 (b F4 e~.

>.

on (D *4 z~.

C>\J S)

TABLE OF DIFFERENCES FOR PAIRS OF VOWELS
BONE CONDUCTION

27
TABLE V--Continued

IN BANDWIDTH I

 

9.6
7.
7.
7.

7.

11.

9.6

13.3*
11.6%
11.5*
11. a
11.0*
13.5*
15.0*
16.3*
13.3%
13.3%

C

   
  

AC

TC

TABLE OF

DIFFERENCES FOR PAIRS OF VOWELS IN

28

TABLE V--Continued

TISSUE CONDUCTION

 

A.

I:

€5

-él---

'32

CCLI

.9

()

BANDWIDTH I

 

Ll

/\

SET

 

4.4%

903*

8.7*

13.1*

14.2*

13.8*

11.9*

706*

4.2*

11.2*

8.2*~L

 

3.8*

806*

8.1*

12.5*

13.6*

13.2*

11.3*

7.0*

3.5*

10.5*

7.6%

 

202*

702*

6.6*

11.0*

12.1*

12.0*

9.8*

5.5*

2.0*

900*

6.1*

 

.5

5.4*

4.8*

9.2*

10.3*

9.9*

8.0*

3.7%

7.2*

4.3*

 

36 Flfb t4 bu

.4

5.3*

4.8*

9.2*

10.3*

.9

8.0*

307*

7.1%

4.2*

 

.4

5.2*

4.6*

9.1*

10.2*

9.7*

8.9*

306*

6.6*

3.7*

 

4.8*

4.2*

8.6*

9.7*

9.3*

7.4%

3.2*

9.2*

6.2*

 

2.4*

7.3*

6.7*

11.1*

12.2*

11.8*

9.9*

5.6*

10.6*

7.8*

 

3.9*

8.8*

8.2*

12.6*

13.7*

13.3*

11.4*

7.2*

11.9*

9.0*

 

5.2*

10.1*

9.5*

13.9*

15.0*

14.6*

12.7*

8.4%

9.0*

6.0%

 

2.3*

7.2*

6.6*

11.0*

12.1*

11.7*

9.8*

5.5*

9.0*

6.1*

 

2.3*

7.2*

6.6*

11.0*

12.1*

11.7*

9.8*

5.5*

9.0*

6.1*

 

 

7.4%

2.5*

3.1*

1.3*

2.4*

2.0*

.1

4.2*

.7

3.6%

 

11.1%

6.2*

6.8*

2.4*

1.3*

1.7*

3.6*

7.9*

4.4*

7.3%

 

10.6*

5.7*

6.3*

1.9*

.8

1.2*

3.1%

7.4*

3.9*

6.8*

 

14.6*

9.7*

10.3*

5.9*

4.8*

5.2*

7.1*

11.4*

7.9*

10.8%

 

6.0*

11.1*

11.7*

7.3*

6.2*

6.6*

8.5*

12.8*

.3*

12.2%

 

i150O—X

10.1*

10.7*

6.3*

5.2*

5.6*

7.5*

11.81"r

8.3*

11.2*

 

14.1*

9.2*

9.8*

5.4*

4.3*

4.7*

6.6*

10.9*

7.4*

10.3P'

 

10.8*

5.9*

6.5*

2.1*

'1.0

1.4*

3.3*

7.6%

4.1*

7.0*

 

805*

3.6*

4.2*

.2

1.3*

.9

1.0

5.3*

1.8*

4.7%

 

7.7%’

2.8*

3.4*

100 7

2.1*

1.7*

.2

4.5*

1.0

3.9*

 

13.3*

8.4*

9.0*

4.6*

4.5*

3.9*

5.8*

10.1*

606*

9.5*

 

11.3*

6.4*

7.0*

2.6*

2.5*

1.9*

3.8*

8.1%

4.6*

7.5%

 

 

4.9*

4.3*

8.7*

9.8*

9.4%

7.5*

3.2%

6.7*

3.8*

 

.6

3.8*

4.9*

4.5*

2.6*

1.7*

1.8*

1.?6

 

4.4*

5.5*

5.1*

3.2*

1.1

2.4%

.5

 

1.1*

.7

1.2*

5.5%

4.9*

 

.4

2.3*

6.6%

3.1*

6.0%

 

1.9*

6.2*

2.7%

5.6-2?

 

4.3*

.8

3 . 7% 1

 

’2‘ I:
w'0J

.6

 

4.7%

(1")

 

70C3“

 

MIST—I
O

\f) H
x v

 

 

0%.:> I: <3 c) \) x) 5% (rich #4 r~.cg,)> x: :3 c) \3 5) 2310) n» +4 e-hi 2» x: :2 C) \J

 

 

 

 

 

 

 

 

 

 

 

 

 

 

TABLE VI

TABLE OF DIFFERENCES FOR PAIRS OF VOWELS IN BANDWIDTH II

Xmme.(-l>:

\)

AIR CONDUCTION

 

With a critical difference of 4.12 at the 5 percen

level of confidence, the values for the pairs of
vowels marked with an asterisk show a significant

 

TABLE VI--Continued

i TABLE OF DIFFERENCES FOR PAIRS OF VOWELS IN BANDWIDTH II
BONE CONDUCTION

 

 

.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

IIeEaeap'ov A3
.3 .3 1.4 5.7% 8.9% 8.1% .7 .6 2.0 212 1 5.3% .8
4.6% 4.6% 3.5 10.6% 13.8% 13.0% 11.9% 4.3% 2.9 2.7 10.2% 5.4%
-4.2% 4.2% 3.1 10.2% 13.4% 12.6% 11.5% 3.9 2.5 2.3 9.8% 5.0%
.4 .4 1.5 5.6% 8.8% 8.0% 6.9% .7 2.1 2.3 5.2% .4
2.4 2.4 3.5 3.6 6.8% 6.0% 4.9% 2.7 4.1% 4.3% 3.2% 1.6
3.4 3.4 4.5% 2.6 5.8% 5.0% 3.9 3.7 5.1% 5.3% 2.2 2'91.
1.3 1.3 2.4 4.7% 7.9% 7.1% 6.0% 1.6 3.0 3.2 4.3% .5
5.5% 5.5% 4.4% 11.5% 14.7% 13.9% 12.8% 5.2% 3.8 3.6 11.1% 6.3%
6.9% 6.9% 5.8% 12.9% 16.1% 15.3% 14.2% 6.6% 5.2% 5.0% 12.5% 7.7%
4.6% 4.6% 3.5 10.6% 13.8% 13.0% 11.9% 4.3% 2.9 2.7 10.2% 5.4%
9.6% .4 1.5 5.6% 8.8% 8.0% 6.9% .7 2.1 2.3 5.2% .4
4.6% 4.6% 3.5 10.6 13.8% 13.0% 11.9% 4.3% 2.9 2.7 10.2% 5.4%
0 1.1 6.0% 9.2% 8.4% 7.3% .3 1.7 1.9 5.6% .8
1.1 6.0% 9.2% 8.4% 7.3% .3 1.7 1.9 5.6% .8
7.1% 10.3% 9.5% 8.4% .8 .6 .8 6.7% 1.9
3.2 2.4 1.3 6.3% 7.7% 7.9% .4 5.2%
.8 1.9 9.5%10.9% 11.1% 3.6’ 8.4%
1.1 8.7%10.1% 10.3% 2.8 7.6%
7.6% 9.0% 9.2% 2.7 6.5%
1.4 1.6 5.9% 1.1
.2 7.3% 2.5
.5% 2.7
4.8%

 

 

 

 

 

 

 

 

_.__...1._—,_

 

 

 

 

 

 

..,.‘y..-- '—4 ———

 

.—.q—-- "qt—v-_‘_~ p-44— “.4 N‘. ._.._

 

 

_ 8'5”
03>1:4ovngmo‘Hh.m>;cgOcpxmer.Q>gc}oopgmmHz~.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

AC

BC

TC

u;>:<ovpxmmH+.ui>cc:ovpgmmHh.ui>cc:ooe%mmHs.

TABLE OF DIFFERENCESFOR PAIRS OF VOWELS

TABLE VI--Continued

31

IN BANDWIDTH II TISSUE CONDUCTION

 

 

.‘vvv- vv00--. —~'.0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LIeEaeQ70‘v‘LAA3‘
1.8 2.4 1.9 9.3% 1.5 .8 10.3 3.1 0 .1 .9.Z*1_2.6
6.7% 7.3% 6.8% 4.2% 16.4% 15.7% 15.2 8.0% 4.9% 4.8% 14.6% 7.5%
6.3% 6.9% 6.4% 13.8% 16.0* 15.3% 14.84 7.6% 4.5% 4.4% 14.2% 7.1%
1.7 2.3 1.8 9.2% 11.4% 10.7% 10.24 3.0 .1 .2 9.6% 2.5

.3 .3 .2 7.2% 9.4% 8.7% 8.2% 1.0 2.1 2.2 7.6% .5
1.3 .7 1.2 6.2% 8.4% 7.7% 7.2% 0 3.1 3.2 6.6% .5

.8 1.4 .9 8.3% 9.5% 9.8% 9.34 2.1 1.0 1.1 8.7% 1.6
7.6% 8.2% 7.7% 15.1% 17.3% 16.6% 16.1 8.9% 5.8% 5.7% 15.5% 8.4%
9.0% 9.6% 9.1% 6.5% 18.7% 18.0% 17.5% 10.3% 7.2% 7.1% 16.9% 9.8%
6.7% 7.3% 6.8% 14.2% 16.4% 15.7% 15.2% 8.0% 4.9% 4.8% 14.6% 7.5%
1.7 2.3 1.8 9.2% 11.4% .7 10.2% 3.0 .1 .2 9.6% 2.5
6.7% 7.3% 6.8% 14.2% 16.4% 16.7* 15.2% 8.0% 4.9% 4.8% 14.6% 7.5%
2.1 2.7 2.2 9.6% 11.8% 11.1% 10.6 3.4 .3 .2 10.0% 2.9
2.1 2.7 2.2 9.6% 11.8% 11.1% 10.6% 3.4 .3 .2 10.0% 2.9
3.2 3.8 3.3 10.7% 12.9% 12.2% 11.7% 4.5% 1.4 1.3 11.1% 4.0
3.9 3.3 3.8 3.6 5.8% 5.1% 4.6% 2.6 5.7% 5.8% 4.0 3.1
7.1% 6.5% 7.0% .4 2.6 1.9 1.4 5.8% 8.9% 9.0% .8 6.3%
6.3% 5.7% 6.2% 1.2 3.4 2.7 2.2 5.0% 8.1% 8.2% 1.6 5.5%
5.2% 4.6% 5.1% 2.3 4.5% 3.8 3.3 3.9 7.0% 7.1% 2.7 4.4%
2.4 3.0 2.5 9.9% 12.1% 11.4% 10.9% 3.7 .6 .5 10.3% 3.2
3.8 4.4% 3.9 11.3% 13.5% 12.8% 12.3% 5.1% 2.0 1.9 11.7% 4.6%
4.0 4.6% 4.1% 11.5% 13.7% 13.0% 12.5% 5.3% 2.2 2.1 11.0% 4.8*b
3.5 2.9 "3.4 4.0 6.2% 5.5% 5.0% 2.2 5.3% 5.4% 4.4% 2.7
-1.3 1.9 1.4 8.8% (1.0% 10.3% 9.8% 2.6 9.5% 9.4% 9.2% 2.1

.6 .1 7.5% 9.7% 9.0% 8.5% 1.3 1.8 1.9 7.9% .5
.5 6.9% 9.1% 8.4% 7.9% .7 2.4 2.5 7.3% .2
7.4% 9.6% 8.9% 8.4% 1.2 1.9 2.0 7.8% .7
2.2 1.5 1.0 6.2% .7 .6 .4 6.7%
.7. 1.2 8.4% 11.5% 11.6% 1.8 8.9%
.5 7.7% 10.8% 10.9% 1.1 8.2%
7.2% 10.3% 19.4% .6 7.7%
3.1 3.2 6.~% .5
.1 7.7% 2.6
f.€t 2.7
7.1%
I

 

 

 

sl—

TABLE VII

TABLE OF DIFFERENCES FOR PAIRS OF VOWELS IN BANDWIDTH III
AIR CONDUCTION

    

e c ' "U‘ u
12. 11.6 22.1% 25.1% 29.4% 30.3% 22. 18.6% 12.2%

1.2 . 9.4 1203* 1607* 1706*. 9.6 509* .5
10.6 14.5* 17.9* 18.8* 10.8 7.1* .7

2.9 7.3% 8.2% .2 3.5 9.9%
4.4 5.3 2.7 7.4% 12.8%

9.1% 7.1 10.8% 17.2%

8.0% 11.7% 18.1%

3.7 10.1*
6.4*

AC

BC With a critical difference of 5.94 at the 5 percent

level of confidence, values for the pairs of vowels

C) x: :3 a; r» (a +4 >¥.9i :> s: c: <> \3 ;> 96 no (9 b4 .4

that are marked with an asterisk show a significant

difference.

GFJ rm 03 >»

H
C)
>C§Ovp¥m

33
TABLE VII--Continued

TABLE OF DIFFERENCES FOR PAIRS OF VOWELS IN BANDWIDTH III
" BONE CONDUCTION

 

8.6% 8.1% 15.1% 16.5% 20.4% 22.3% 18,29 13,09

AC

BC

H
O
s>c<£ox>m>smesut>z

42119 +4 an

99

OvnxmmHhJa>

4.1
2.9
13.5*
16.4*
20.8*
21.7*
13.7*
10.0*
3.6
19.2*
10.9*

8.8*

4.7
3.5
14.1*
17.0*
21.4*
22.3*
14.3*
10.6*
4.2
19.8*
11.5*
8.2*
.6

2.3
3.5
7.0*
9.9*
14.3*
15.2*
7.2*
3.5
2.9
12.7*
4.5
5.3*
6.5*
7.0*

3.8
5.0
5.6
8.5*
12.9*
13.8*
5.9*
2.1
4.3
11.4*
3.0
l . *
7.9*
8.5*
1.5

7.6*
8.8*
1.8
4.7
9.1*
10.0*
2.0*
1.7
8.1*
7.6*
.8
20.5“)"
11.7*
12.3*
5.3

3.8

9.5*
10.7*
.1
2.8
7.2*
8.1*
.1
3.6
10.0*
5.7
2.7
’2.4
13.6*
14.2*
7.2*
5.7
1.9

5.5
6.7*
3.9
6.8*
11.2%

I" 1201*

4.1
.4
6.0*
9.6%

1.3

.2
U0

9.6*
10.2*
3.1
1.7
2.1
4.0
1.3
5.0

9.8*

.3
1.5
9.1*

12.0*
16.4*
17.3*

.3*
5.6

.8

 

TABLE OF DIFFERENCES FOR PAIRS OF VOWELS IN BANDWIDTH III

TABLE VII--Continued

TISSUE CONDUCTION

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1: 3C es 5? egg CL 5? c1 .4
6.4% 4.8 5.5 3.1 2.8 .1 2.3 1.6 4.1
19.1% 17.6% 18.3% 15.8% 15.5% 12.7% 10.4% 11.2% 12.9%
17.9% 16.4% 17.1% 14.7% 14.4% 11.5% 9.2% 10.0% 1.7%
28.5% 27.0% 17.7% 25.2% 24.9% 22.1% 19.8% 10.6% 22.3%
31.4% 29.9% 30.6% 28.2% 27.8% 25.0% 22.7% 3.5% 25.2%
35.8% 34.3% 35.0% 32.5% 32.2% 29.4% 27.1% 27.9% 29.6%
36.7% 35.2% 35.9% 33.4% 33.1% 30.3% 28.0% 28.8% 29.5%
28.7% 27.2% 27.9% 25.4% 25.1% 22.0% 20.0% 20.8% 22.5%
25.0% 23.5% 24.2% 21.7% 21.4% 18.6% 16.3% 17.1% 18.8%
8.6% 17.1% 17.8% 15.3% 15.0% 12.2% 9.9% 10.7% 12.4%
34.2% 32.7% 33.4% 30.9% 30.6% 27.8% 25.5% 26.3% 28.0%

25.9% 24.4% 25.1% 22.7% 22.4% 19.5% 17.2% 18.0% 19.7% -
6.2% 4.7 5.3 2.9 2.6 .2 2.5 1.7 0
15.0% 13.5% 14.2% 11.7% 11.4% 8.6% 6.3% 7.1% 8.8%
14.4% 12.9% 13.6% 11.2% 10.8% 8.0% 5.7 7.5% 8.2%
21.5% 19.9% 20.6% 18.2% 17.9% 15.0% 12.8% 13.5% 15.2%
22.9% 21.4% 22.1% 19.7% 19.4% 16.5% 14.2% 15.0% 16.7%
26.7% 25.2% 25.9% 3.5% 23.2% 20.3% 18.0% 18.8% 20.5%
28.6% 27.1% 27.8% 25.3% 25.1% 22.2% 19.9% 20.7% 22.4%
23.3% 21.8% 22.5% 20.1% 19.7% 16.9% 14.6% 15.4% 17.1%
19.6% 18.1% 18.8% 16.3% 16.0% 13.2% 10.9% 11.7% 13.4%
14.8% 13.3% 14.0% 11.6% 11.3% 8.4% 6.1% 6.9% 8.6%
24.6% 23.1% 23.8% 21.3% 21.0% 18.2% 15.9% 16.7% 19.4%
19.4% 17.9% 18.6% 16.1% 15.8% 13.0% 10.7% 11.5% 13.2%

1.5 .8 3.3 3.6 6.4 8.7% 7.9% 6,2.
.7 1.7 2.0 4.9 7.2% 6.4% 4.7
2.4 2.7 5.6 7.9% 7.1% 5.4
.3 3.2 5.4 4.7 2.9
2.8 5.1 4.3 2.7
2.3 1.5 3 .2
.8 5 8 2.5
5 1 1.7
2 r 1.4
3.4

 

 

 

%>chvD%mmHe-N>cq’oupxmmI—qe.w>:dovaSmwt-Hn

 

 

 

 

 

 

 

 

 

 

 

 

 

 

35

When comparing all of the front vowels [(1,]: 9,8, 3Q]
with each other over all three bandwidths, approximately one-
third tend to be alike in mean energy. That is, there is not
enough difference between this one-third to be significant
according to the critical difference determined for each band-
width. Approximately one—fourth of the back vowels[1f ,LA,C2,
7 ], and one—sixth of the central vowels[/\ ,.3‘]tend to be
alike in mean energy over all three bandwidths. When looking
at the bandwidths separately, bandwidth II contains the great-
est number of vowel pairs that tend to be alike in mean energy
or that are not significantly different in mean energy. Band-
width III has the second highest number with bandwidth I having
the fewest. In all three bandwidths, the front vowel pairs
are the most numerous in tending to be alike in mean energy to
the extent that they are not significantly different with the
back vowel pairs being second highest in number.

When comparing all of the pairs of front vowels with
each other, it can be seen that ten pairs of front vowels are
alike in mean energy in bandwidth I, seventeen pairs in band-
width II, and one pair in bandwidth III. In bandwidth I there
are four pairs of central vowels alike in mean energy. The
front vowels are the only pairs of vowels with the exception
of the central vowels in bandwidth I that show a likeness in
Inean energy. None of these pairs of vowels is the same from

bandwidth to bandwidth.

36

More research is needed in this area to determine the
significance of more pairs of front vowels falling within
bandwidth II, and to determine the significance of more
pairs of front vowels being alike than the back vowels,
or central vowels.

To find the answer to the fourth question concerning
the relationship between any two points of pickup and the
amount of energy for any given vowel, correlation coef-
ficients between each possible pair of pickup points for each
of the twelve vowels were obtained using the Pearson Product-
Moment correlation coefficient.

The results of this treatment are presented in Table
VIII.

An attempt was made to classify the significant coef-
ficients according to front, back and central vowels. It
can be seen that seven of the fifteen significant coefficients
fall among the front vowels, six among the back vowels, and
two among the central vowels. It should also be noted that
eleven of the significant coefficients fall in bandwidth
I among all combinations.

The vowels[€ , 0 , Vjchow a significant relationship
"between AC-BC and AC-TC in bandwidth I with values approach-
ing significance between BC-TC.

An explanation of these findings cannot be made with-

cmxt further research in this area.

AC—BC

AC-TC

BC-TC

II
III

II
III

II
III

37

TABLE VIII

PRESENTATION OF CORRELATION COEFFICIENTS BETWEEI ALL
COMBINATIONS OF TWO POINTS OF PICKUP FOR ALL TWELVE
VOWELS IN THE THREE BANDWIDTHS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

II I e 8 58 a 7 o v u A j"
.73 .95% .57 .97% .37 .03 .57 .95-4— .igifl‘ .517 .89 ' .67:
-.36 -.21 .21 .22 .40 .23 .42 .09 .30 .30 ., -.002
.82* .73 .74 .32 .41 .36 .41 .71 .48 .01 -.29 .30
.77 .48 .90* .83* .57 -.58 .14 .84* .91* .14 .75 .67
.87* .65 .48 -.14 .06 .61 .49 .40 -.O6 .87* .63 .82*
.353 .13 .75 .59 .38 .32 .49 .59 .24 .24 -.12 .68
.81* .37 .46 .78 -.001 -.08 .06 .68 .64 .71 .66 .76
-.O7 .34 .55 .35 .24 .78 .37 .39 .38 .31 .34 .40
.47 .30 .67 .48 .53 .25 .36 .54 .42 .09 .05 A66

 

 

 

 

 

 

 

 

 

 

 

 

 

 

*In a sample of this size with df=4, a coefficient must be equal to
or greater than i .81 to be significant at the 5 percent level of
confidence.

 

38

Therefore, the fourth hypothesis which states that
there is no significant correlation between the amount of
energy for any given vowel and any two points of pickup

would be accepted, in general.

Discussion

 

The purpose of this study was to determine the signifi—
cance of energy variations of twelve vowel sounds using
three methods of pickup within three different bandwidths.

One can see from Table I that the mean vowel ampli-
tude is greater for air conduction than for bone or tissue
conduction in all three bandwidths. Bandwidths II and III
follow the expected pattern of having bone conduction
second highest and tissue conduction third since bone is
thought to be a better conductor of sound waves than tissue
of unequal density. However, bandwidth I does not follow
this pattern. Further study is needed in this area to estab-
lish the reasons behind these findings.

The results of data analysis indicated a significant
difference in the mean amount of energy between vowels and
between pickup methods at the 5 percent level of confidence
for all three bandwidths. There was a significant inter—
action between vowels and points of pickup at the 5 percent
level of confidence for bandwidths I and III but not for

bandwidth II.

39

Since it was desirable to know which vowel sounds
were different from which other vowel sounds, the critical
difference was computed for each bandwidth and the resulting
information was presented in Tables V, VI, and VII.

Using these tables it can be noted that in comparing
the two vowel sounds[e hnd[O],the significant differences
in bandwidths I and II are a function of the pickup points.
This can be explained by the fact that there was no signifi-
cant difference between these two vowels when they were
picked up by the same method. However, there was a signifi-
cant difference between bone and tissue pickup in band—
width II. In bandwidth III there was a significant dif-
ference between these two vowels when each of them was picked
up by the same method but there was no significant difference
between the two when one was picked up by air condictuon
and the other by bone conduction. This information indicates
that there must be some interaction taking place between
these two vowel sounds and the points of pickup.

When comparing the two vowel soundsfixabndflk],it can
be seen that there is some interaction taking place in all
three bandwidths. In bandwidths I and II there was a
significant difference between the two vowel sounds when
they were picked up by the same method but there was no
significant difference between them when one was picked up
by air conduction and the other by tissue conduction. In

bandwidth III there was a significant difference in all

4O

combinations of the two vowel sounds except when they
were both picked up through tissue conduction.

In looking at the two vowel sounds[/\]and[U]it is seen
that some interaction is evident in bandwidths I and III
but the significant differences in bandwidth II are a '
function of pickup points.

In comparing all of the pairs of front vowels with
each other, it is noted that most of the like pairs fall in
bandwidth I with the second highest number in bandwidth II.
However, none of these pairs of vowels is the same from
bandwidth to bandwidth.

In terms of significance, the front vowels include
the greatest number of pairs that tend to be alike in mean
energy, that is they are not significantly different from
each other, with the back vowels second in line followed by
the central vowels.

In addition to the above information, it was desirable
to attempt to find the relationship, if any, between any
two points of pickup and the amount of energy for any given
vowel. From the correlation coefficients between the three
points of pickup and each of the twelve vowel sounds, the
significant coefficients were found to be located mostly
in bandwidth I. The vowel sounds[€ ,O, 'U‘]show a signifi-
cant relationship between AC-BC and AC-TC in bandwidth I

with values approaching significance between BC—TC.

CHAPTER V
SUMMARY AND CONCLUSIONS

Summary

During the past thirty-six years investigators have
designed experiments to measure the intelligibility of
tissue transmitted signals and to record vocal signals from
various locations on the body of Speakers.

From past studies it is known that intelligible speech
can be picked up at various locations on a speaker's body
other than the lips.

The purpose of this study was to compare the energy
variation at each of three bandwidths over twelve vowels of
the English language simultaneously recorded at the lips and
forehead of six different Speakers using three types of
pickup devices. This information provides some insight
into the possibility of utilizing the forehead as a point
of pickup.

The three bandwidths involved in this study were
expressed in cycles per second and divided into groupings
of 100—199, 200—499, and 500-999 and were referred to as

bandwidths I, II, and III respectively. The twelve vowel
sounds that were used were[¢ ,I,e,€,a€,a, 2,0,7!) U,/\,

and.3~].

41

42

A separate analysis of variance was done-on-relative
intensity measures derived for each of the three-bandwidths.
The analysis for bandwidth I showed a significant difference
in the mean amount of energy between vowels, between pickup
methods, and a significant interaction between vowels and
points of pickup at the 5 percent level of confidence.

Since the F—Test revealed whether the amount of
energy for any one or more of the twelve vowel sounds was
significantly different from the amount of energy for any
of the other vowels, a test of individual comparisons was
desirable to determine which pairs of vowels were signifi-
cantly different in each of the three bandwidths over all
three methods of pickup. Critical difference scores were
computed for each bandwidth with the results given in
Tables V, VI, and VII.

By using Tables V, VI, any VII and two vowel sounds
can be compared. It can be established whether the signifi-
cant differences are a function of the pickup points or
whether there is some interaction taking place between the
two vowel sounds in question and the points of pickup.

Correlation coefficients between the three points of
pickup and each of the twelve vowel sounds were obtained
using the Pearson Product-Moment correlation to determine
the relationship between any two points of pickup and the
amount of energy for any given vowel. The results of this

treatment are given in Table VIII.

11,3

Conclusions

 

Upon inspection of the results obtained from data

analysis the following conclusions can be made:

I.

I’D

10.

The mean vowel amplitude is greater for air conduction
than for bone or tissue conduction in all three bandwidths.
The mean amount of energy differs from vowel to vowel
within each of the three bandwidths.

The mean amount of energy differs as a function of pick%
up methods within each of the three bandwidths.

There is an interaction between vowels and points of pick-
up within bandwidths I and III but not within bandwidth II.
The front vowels tend to be more alike in mean energy over
all bandwidths than the back or central vowels.

Bandwidth II contains the greatest number of vowel pairs
that are not significantly different in mean energy.

The front vowel pairs are the only pairs of vowels that
show an exact likeness in mean energy in bandwidths II

and III. Bandwidth I also has four pairs of central vow-
els showing an exact likeness in mean energy.

Eleven of the fifteen significant coefficients fall in
bandwidth I among all combinations.

The vowels [E:,<3,'U1 show a significant relationship
between AC—BC and AC—TC in bandwidth I with values ap—
proaching significance between BC-TC.

In comparing two vowel sounds, some show interaction

taking place between them and the points of pickup as

44

in [32] and [[1], and some show that the significant
differences are a function of pickup points such as

in [/\] and [LA].

Implications for Future Research

 

This study suggests several areas for continued research.
Since air conduction did show a greater mean vowel amplitude
in all of the three bandwidths, further study could be done
to test this trend. Would this trend be evident if another
study were done using different speakers or using speech sig-
nals instead of vocal signals? Would bandwidth I still show
tissue conduction pickup as having a higher mean vowel ampli-
tude than bone conduction pickup? Would the trend of fewer
alike vowel pairs falling in bandwidth I be continued? Would
the front vowels have more alike pairs than the back or cen-
tral vowels?

This study showed eleven of the fifteen significant
coefficients falling in bandwidth I among all the combina-
tions. Would other points of pickup show similar results?

Since the condenser microphone that was used on the
forehead for tissue conduction pickup was of the_same type
that was used in front of the lips for air conduction pickup,
the fact that this type microphone would produce better re—
sults with air conduction pickup is evident. If this study
were done again using equipment especially made for tissue
pickup, would tissue pickup still produce a higher mean vow—
el amplitude than bone conduction pickup as was true for

bandwidth I in this study?

45

APPENDIX I

PRESENTATION OF RAW SCORES FOR SPEAKER I
OVER TWELVE VOWELS IN SIX BANDWIDTHS
FOR AIR, BONE, AND TISSUE
CONDUCTION PICKUP

 

CPS .4 IE e: E: 38.. CL 7 c> 1r L1 .0 3?

pa nah-

 

 

I 100-199 31 84 29.528 28 29 30 81.5 31 33 35.533.
II p 200-499 33.337.7 36.328.7 26 29.7 30.337 35 38 33

 

III 500-999 9 20.8 20.231.2 30 34.2 34 27.6 23 20.6 35 29.4
IV 1000-1999 3 16 17 21 20.524 21 14 11 10.5 24.529.5

AC

 

 

V 2000-4999 13.512.7 14.713.3 12.3 6 5 O O O 11.3 0
VI 5000-9999 0 O 0 O O O O O 0 O O

 

 

 

 

CO
I 0'
)\

I 100-199 24 20.5 24 3 20 -1 22.5.9 2. 3 21.522.5

 

II 200-499 26.729.3 32.723.3 24.3 2 23 30.3 29.332.3 24.333.3
III 500-999 6 17 1 .625.6 24.8_6.4 27.624.2 19 16.8 25.224.4

 

 

IV 1000-1999 2.5 9.5 16.5 19 U9 15 O 9.5 8 19 24.5

 

 

6

8
7 V 2000-4999 0 O 0 O '5.7 O O O O 0 O 1.7
VI 5000-9999 0 O 0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 0 0 0 0 0 0 0
I 100-199 23 13 23 19 22 20.5 20.521 22 84 23 26

II 200-499 31.332 32.723.3 23 23 22.332.3 35.336 23.734.3_

III 500-999 0 5.8 6.2 8.8 7 8.4 9.2 3.2 6.6 4.6 9.813.4
TC Iv 1000-1999 0 0 0 0 0 0 0 c 0 0 0
v 2000-4999 0 o 0 0 0 0 0 0 0
VI. 5000-9999 0 o 0 0 0 0 o 0 0

 

 

 

 

46

APPENDIX I--Continued

PRESENTATION OF RAW SCORES FOR SPEAKER II
OVER TWELVE VOWELS IN SIX BANDWIDTHS
FOR AIR, BONE, AND TISSUE CONDUCTION

II

III
AC Iv

VI

II

III
BC

VI

II

TC III

VI

  

CS

 

lOO-l99
200-499

1000-1999
2000-4999
5000-9999

100-199
200-499
500-999

1000-1999
2000-4999
5000-9999

loo-199

200-499

500-999
1000-1999

2000-4999
5000-9999

IPICKDI’

h-

 

47

APPENDIX I--Continued

PRESENTATION OF RAW SCORES FOR SPEAKER III
OVER TWELVE VOWELS IN SIX BANDWIDTHS
FOR AIR, BONE, AND TISSUE CONDUCTION

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

‘ PIIHGJP
cps~ l: I]? 65 §;, :81 c2. :9 C) '7!’ L1 /\ g:;_
I 100-199 34 ,9.5 29 26.5 75.5 24.593.5 26 28 28.5 24 4.5
II 200-499 32.740.3 38.784 32.3 32 33 40.741.737 31 38
III 500-999 10.223 23 32.4 36.4 40.441.8 31.431 23.8 38.429
AC IV 1000-1999 11 25.5 26 32 34 21 32.5 18 23 17 30 80.5
V 2000-4999 21 28.3 28 30 34 20.325.7 15.716.7 9.3 31 10.4
VI 5000-9999 1 5.2 6.8 8.4 7.2 2.6 0 0 3 0 2.8 0
I loo-199 14 13.5 13.513 9.5 11 Io 13 14 7 11 24
II 200-499 32 29.7 28.324 19 19.324 28.328.729.3 30 27.3
BC III 500-999 10.411.4 12.818 20 23. 25.2 19.217.815 22.415.4
IV 1000-1999 3 6 5.510 15 11 12.5 7. 9.5 7 9.510
V 2000-4999 0 0 0 0 '0 0 0 0 0 0 0
VI 5000-9999 0 0 0 0 0 0 0 0 0 0 0
I- 100-199 21 28.5 25.5 23.520.520.5 21 25 -6 9 20.5 4.5
II 200-499 32 81.3 28.7 23 19.319 21.3 26.727.329 20.726
III 500-999 10.6 8 4.6 5.8 9.614.6 5.2 9.6 7.8 9.8 14.2 6
. TC IV 1000-1999 0 3.5 0 0 1.5 0 0 0 0 0 0
V 2000-4999 0 ‘ 0 0 0 0 o 0 0 0 0
VI. 5000-9999 0 0 0 0 O 0 0 0 0 0 0

 

 

 

 

48

APPENDIX I--Continued

PRESENTATION OF RAW SCORES FOR SPEAKER IV
OVER TWELVE VOWELS IN SIX BANDWIDTHS
. FOR AIR, BONE, AND TISSUE CONDUCTION
PICKUP

I 100-199
II ‘200-499
III 500-999
IV 1000-1999
V 2000-4999
VI 5000-9999

AC

I lOO-199

II 200-499

III 500-999
IV lOCO-1999
V 2000-4999
VI 5000-9999

BC

I 100-199
II 200-499
III 00- 9

TC Iv 1000-1999
' V 2000-4999
VI 5000-9999

 

II

III
AC 1N

VI

II

III
BC

VI

II

III
TC

VI

49

APPENDIX I--C0ntinued

PRESENTATION OF RAW SCORES FOR SPEAKER V
OVER TWELVE VOWELS IN SIX BANDWIDTHS
FOR AIR, BONE, AND TISSUE CONDUCTION

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PIIUGJP

cps 4 It a! {3- 53 GE 7’ ‘24—1f {4 43 i££_.

100-199 37.5 8 36 37 37 33 33 38 40 40.5 36.5 34

200-499 29.335 36 33 29.3 25 29.7 36 38 36.7 31.7 34

500-999 10.825 21 31 36.2 40.240.8 32.831.8 22 37.8 24.8
1000-1999 8.532 20 27 27 31 26 21.521 9.5 23 26
2000-4999 20 25.7 24 25.728 21.:13.7 13 7.7 2 16.3 9.7
5000-9999 0 4.6 4.6 5.6 7.6 0 0 2.6 0 0 0 0
: 100-199 28 21 20.518.518.5 13 17 23 26 26.5 17 20.5
, 200-499 34 83 35.730 26 20.324 34.738.3 39.3 26 34.3
. 500-999 l3.419.6 18 22 26 24.828.4 26.227 20.6 24.8 18.6
1000-1999 7.512 9 14.515.5 14.513 10.518 6.5 11 14
2000-4999 0 0 :0 0 0 o O 0 0 0 0 0 I
5000-9999 0 0. 0 0 0 0 0 0 0 0 0 0

100-199 38 29 29.5 26.527 18 22.5 30.5 33 36.5 22 26.5
* 200-499 24.323.326 19.717 12 15.7 24.3 28.729.3 18.3 24.7

500-999 0 2 2 1.4 4.6 3.4 7.2 5.6 5.4 3.4 2.6 0
1000-1999 0 0 O 0 0 o o O O 0
2000-4999 0 0 0 0 0 0 0 0 0 0 0
5000-9999 , 0 ‘0 0 0 0 0 0 0 0 c 0

 

 

 

 

II

III
AC IV

VI

II

III
BC

VI

II

III
TC

VI

50

APPENDIX I--Continued

PRESENTATION OF RAW SCORES FOR SPEAKER VI
OVER TWELVE VOWELS IN SIX BANDWIDTHS
FOR AIR, BONE, AND TISSUE CONDUCTION

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PICNRH?
CPS .6 If g; £3 49 CL. .7 92. 1f’ L1 24 if
100-199 39 38 35 33.5 32.5 33 34.5 38 38.5 40 35 36
200-499 32.737.7 39 32.3 27 28.731.7 40.341.7 38.733.7 39.3
500-999 11.624.2 26.232.6 40.2 44.446 39.431 22 37.2 33.4
1000—1999 5.516 17.526 33.5 31.531.5 26.524 10 25 33.5
2000-4999 25.728 29 29.7 28 25.324.7 17.315.7 8.719.7 12.3
5000-9999 2.8 5 6. 4 3.8 5.6 5.2 2 0 0 0 0
ICC-199 30' 25.5 22 16 8 11.5 14 22 26 28 17 20.5
200-499 35.738.7 39.731 19 0.3 24.338.3 40.739.3 29.3 32.7
500-999 9.6 9.4 20.821.2 21.826.6 30.230.5 24.419.8 26 25.4
1000-1999 2.5 9.5 7 8 15 12.5 14 18 10 11 7 14
.2000-4999 0 0 0 0 0 0 o 0 c 0 0 0 '
5000-9999 0 o O 0 0 0 0 0 0 0 0 0
100-199 39 33 29 24 14.518.5 20.532. 33.588. 25.5 29
200-499 34.336.3 38 29.3 21 22.3 24 37 39 37.7 27 35.3
500-999 6.813.4 13.614 16 19.6 20.822.4 18 31.8 16.8 17.4
1000-1999 0 2 2 O 7 7 8 6 5 3.5 2 3.5
2000-4999 0 0 0 0 0 0 0 0 0 3 0 0
5000-9999 0 0 0 o 0 0 0 0 0 0 0 o

 

 

 

 

BIBLIOGRAPHY

Bocks

Dixon, Wilfrid J., and Massey Frank J. Jr. Introduction to
Statistical Analysis. New York: McGraw—Hill Book
Company Inc., 1957.

 

 

Gray, Giles W., and Wise, Claude M. The Bases of Speech.
New York: Harper and Brothers, 1959.

 

Lindquist, E. F. Design and Analysis of Experiments in
Psychology and Education. Boston: Houghton Mifflin

COmpany, 1953.

 

 

Stevens, S. S. (ed). Handbook of Experimental Psychology.
New York: John Wiley and Sons Inc., 1951.

Articles and Periodicals

 

Moser, Henry M., and 0yer, Herbert J. "Relative Intensities
of Sounds at Various Anatomical Locations of the Head
and Neck during Phonation of the Vowels," The Journal
of the Acoustical Society of America, 30 (April, 1958).

 

 

Mullendore, James M. "Relative Amplitudes of Sound Vibrations
at Various Body Locations,” Speech Monographs, I6 (1949).

 

0yer, Herbert J. "Relative Intelligibility of Speech Recorded
Simultaneously at the Ear and Mouth,” The Journal of
the Acoustical Society Of America, 27 (Novermber, 1955).

 

 

0yer, Herbert J., Moser, Henry M., and Wolfe, Susan M.
”Relationship Of Phonetic Structure to the Intellig-
ibility of Words Simultaneously Recorded at Ear and
Lips,"lJournal of Speech and Hearing Research, 3
(March, 1960).

Simon, Clarence, and Keller, Franklin. ”An approach to the
Problem of Chest Resonance ” Quarterly Journal of
Speech Education, 13 (19275.

 

 

51

52

Snidecor, John C., Rehman, Irving, and Washburn, David D.
"Speech Pickup by Contact Microphone at Head and Neck
Positions," Journal Of Speech and Hearing Research, 2

(September, 1959).

West, Robert. "The Nature of Vocal Sounds," Quarterly
Journal of Speech Education, 12 (1926).

Wise, Claude M. "Chest Resonnance," Quarterly Journal of

Speech, 18 (1932).

Reports

Hirsh, I. J., and Benson, R. W. "Wright Air DeveIOpment
Center, WADC Technical Report NO. 52, 175," (May

1952).

Moser, Henry M., and Dreher, John J. "Operational Tests of
Miniature MicrOphones and Receivers," (Technical
Report NO. 36), AFCRC TN 56-57, (October, 1956).

Other Sources

 

0yer, Herbert J., Personal interviews, Michigan State
University. July, 1962; January, 1963.

ROOM USE ONLY