SPECTROGRAPHIC ANALYSIS OF INTERSPEAKER AND INTRASPEAKER VARlABlLITIES OF PROFESSIONAL MIMICRY Thesis for the Degree of M. A. MICHIGAN STATE UNIVERSITY MALCOLM ELDON HALL 1975 » u on. _ "31‘; ' if LIBRARY ‘3; Michigan $353 (. UnithY llflll 1W3WIWWIIETHWW L, 3 1293 00670 2256 i ABSTRACT SPECTROGRAPHIC ANALYSIS OF INTERSPEAKER AND INTRASPEAKER VARIABILITIES OF PROFESSIONAL MIMICRY by Malcolm Eldon Hall Six well known celebrities from the motion picture, television and political world were mimicked by a profes- sional mimic known for mimicking over 150 different voices of male and female celebrities. The mimicked voices were compared spectrographically with the natural voice of the mimic and with the natural voices of the celebrities them- selves. It was determined spectrographically that all the speakers' voices including the mimicked voices contained a different mean fundamental frequency. The intraspeaker variability of the mimic was not significant as a whole, when comparing his natural voice with the mimicked voice; however, upper frequency variations were prevalent. Signi- ,ficant interspeaker variability oCcurred throughout the spectra when the mimicked voice was compared to that of the subjects' natural voices; therefore, a significant dif- ference was observed spectrographically among speakers. Aural examination of randomized paired recordings was performed by non-trained listeners. Percentages of correct identifications of speakers (being the same or different) when recorded in a quiet environment was higher than percentages of correct responses obtained with recordings containing a noisy environment. In all cases non-similar contexts yielded higher percentages of right decisions than similar contexts. SPECTROGRAPHIC ANALYSIS OF INTERSPEAKER AND INTRASPEAKER VARIABILITIES OF PROFESSIONAL MIMICRY By Malcolm Eldon Hall A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Audiology and Speech Sciences 1975 Copyright by Malcolm Eldon Hall 1975 Accepted by the faculty of the Department of Audi- ology and Speech Sciences, College of Communication Arts, Michigan State UniverSity, in partial fulfillment of the requirements for the degree of Master of Arts. Thesis Committee: MU. , Director Oscar I. Tosi, Ph.D. I " fl ' \ ,/ ‘TXVS' _/ ,. _.v \. c T ' \~__‘_. \‘ ‘ /\\... :-(. - Q Leo V. Deal, Ph. D. i/r \/+‘ HefbeftJ . Oyj;£/ ii ACKNOWLEDGEMENTS I wish to express my appreciation to Dr. Oscar Tosi for his guidance and support as my thesis director. His valuable assistance in the preparation of this thesis shall never be forgotten. ' Special appreciation goes to Mr. Rich Little, the professional mimic, who recorded for me the mimicked voices, gratis. I also wish to thank Dr. Maurice Crane who pro- vided some of the celebrity recordings from his voice library at Michigan State University. The addresses of the celebrities could not have been acquired if not for Mr. Paul Jenkins, of the National Enquirer, Lantana, Florida. I am appreciative to Dr. Frank Horvath, Department of Criminal Justice, Michigan State University, who gave con- structive criticism toward my thesis structure during his visits to the Michigan State Police Laboratory. I wish to thank those whom I may have inadvertantly overlooked that gave encouragement over the last year and ~a half of thesis study. Most of all, I want to thank my wife, Gwen, for being so understanding about my different moods throughout this thesis and the unqualifying assistance in typing and re- typing the numerous rough drafts. My two sons whom I neglected all too often this past year also deserve a word of thanks and apology. TABLE OF CONTENTS CHAPTER II. III. IV. LIST OF TABLES LIST OF FIGURES. INTRODUCTION . Introduction. Spectrography . Mimicry . REVIEW OF LITERATURE . Voice Disguise. Automatic Speaker Verification. Appellate Court Reversal. DESIGN OF STUDY. Subjects. . Instrumentation . . Fundamental Frequency . Aural Discrimination. ANALYSIS OF DATA AND CONCLUSIONS . Spectrographic Analysis Fundamental Frequency Analysis. Aural Discrimination. . Conclusions . SUMMARY AND DISCUSSION FOR FUTURE RESEARCH . Summary . Conclusions . . Recommendations for Future Research . APPENDICES A. Correspondence with Mimic and Celebrities B. Raw Data of Mimicked Text . . . - C. Raw Data of Aural Discrimination Test . LIST OF REFERENCES . . . . . . . . . . . iv Page vi H WWI-4 39 58 59 LIST OF TABLES TABLE . Page 1. Percentage of decisions (A, B, C) as a function of listening for each general category ofa‘., Q , and y” presentations. . . . . . . . . . . . . 31 2. Collation of listeners' answers into ambient and quiet categories . . . . . . . . . . . . . . . 31 3. Percentage of answers (A, B, C) given by listeners in the categories: ambient noise and quiet. . . . 32 LIST OF FIGURES FIGURE 1. Front View of 700 Series Sound Spectrograph . 2. Comparison Spectrograms 3. 4691 Sound Spectrograph . 4. Display of Acoustic Parameters. 5. Graph and Formula for finding fundamental frequency . 6. Spectra of the six subjects mimicked voices and the professional mimic using his natural voice. 7. Subject number five and the mimic uttering the word /bi/ from the phrase F.0.B. and the word /bi/ in isolation, respectively . . . . . . 8. Mimic and subject number five uttering the phoneme /A / in isolation . 9. Mimic and subject number five uttering the word /bnt/ . 10. Mimic and subject number five uttering the word /a1:/. 11. Subject number five and the mimic uttering the word /neva‘/. 12.. Subject number one and the mimic uttering the word /n£va“/- 13. Mimic and subject number three uttering the word /mAst/. l4. Mimic and subject number six uttering the word /d3an/. 15. Subject number one and the mimic uttering the word /taIm/ . 16. Subject number two and the mimic uttering the word /hi/ . 17. Subject number two and the mimic uttering the word /5rS/. vi Page 16 18 20 21 21 21 21 22 22 23 23 24 24 25 List of Figures (cont'd.) 18. 19. 20. 21. 22. 23. Mimic and subject number three uttering the word /haev/ . . . . . . . . Subject number four and the mimic uttering the word /ar/ . Mimic and subject number six uttering the word /nev97. Subject number six and the mimic uttering the word /a-¢nd/ . Mean fundamental frequency between the mimics natural voice and the mimicked voices of the six subjects. Mean fundamental frequency between the subjects natural voices and their mimicked voices. vii 25 25 26 26 27 28 CHAPTER I INTRODUCTION Voice identification is based on the existence of interspeaker variability (differences between speakers) and on the empirical verification that intraspeaker variability (differences within the same speaker) are less than or dif- ferent from the interspeaker variability. Relevant para- meters that reveal these differences may differ accordingly to the method used; therefore, the parameters chosen in this experiment were necessarily empirical, selected to determine reliable characteristics that can be used to spectrographically identify or eliminate speakers' voices mimicked by a professional mimic. The speaker output spectra carry not only semantic information, but speaker-dependent characteristics as well. The frequency bands of relative higher amplitude of the variable output spectra are called "formants" in the case of vowels. There exists also frequency bands Of relative higher amplitude in the case of consonants. The relative position of the center frequency of these bands, their band widths and relative amplitudes determine phonetic elements, as well as the speaker-dependent characteristics, necessary for identification. However, these parameters are variable not only among different speakers but also within the same speaker uttering the same phonetic 1 elements. Coarticulation is one of the important factors responsible for these variations (Daniloff and Hammarberg, 1973). The general acoustic parameters of speech are time, frequency and intensity distribution within all bands of frequency simultaneously present in the instantaneous speaker output. Comparisons of these general spectral para- meters are the basis of all speaker recognition systems,‘ both subjective and objective. Variation of these spectral parameters depend not only on phonetic content but also on speaker individuality. The problem is that even maintaining the text constant, values of the selected parameters will differ not only among different speakers (interspeaker variability) but also within the same speaker (intraspeaker variability) when different utterances of the same text are compared. During visual examination of these variabilities within the spectrogram the trained examiner looks for the following: mean frequencies and band width of vowel for- mants, gaps and type of vertical striations, slopes of for- mants, duration of similar phonetic elements and plosive gaps, energy distribution of fricatives, plosives and inter- formant spaces (Tosi, et al., 1972)., Additional variabilities, which are introduced by environmental reverberations, noise and distortions of transmission and recording systems, are also found within the spectra which makes the examiner's task even more difficult. Spectrography Such differences can be examined spectrographically; therefore, the spectrograph has long since become an un- questionable aid to acoustic scientists in the study of the speech process. Sound spectrographs differ from most modern spectra analyzers in that they scan a magnetically recorded tape loop repeatedly while a heterodyne analyzer slowly scans the frequency domain, gradually building up a display much like a television raster. A new sound spectrograph produced by Voice Identi> fication, Inc., Series 700 (Figure 1), produces direct analysis from a pre-recorded tape. The instrument uses a tape scanner which repeats any 2.5 second segment of tape for spectrographic analysis and includes an automatic tape advance mode for producing contiguous Spectrograms. The tape loop is held in position by a split drum assembly in which a rotating half-track playback head scans the entire tape loop. Another new development is the use of lKHz calibrate lines which span the entire spectrogram. These lines are marked for only one rotation of the working drum and are then kept as unobtrusive as possible while still offering reference frequency marks in the event that a spectrogram should be cut or folded for word isolation. A comparison is shown in Figure 2 of the results obtained using the same tape recorded signal on two different spectrographs. The spectrogram at the top in Figure 2 was made using an early model 4691 instrument, produced by Voiceprint Laboratories, Inc., (A. J. Presti, 1966) shown in Figure 3, whereas the one at the bottom was made on the series 700, (Figure 1)., It can be readily seen that the same high resolution has been achieved in the series 700. The fundamental reliability of the sound spectrograph for voice identification has been accepted in the forensic sciences. However, unanswered questions regarding inter- speaker and intraspeaker variability in purposely disguised voices, and particularly that of mimiCry, has raised the need for further research to determine whether differences do exist. One purpose of this study was to determine whether or not the voice of a professional mimic did contain identi- fiable interspeaker variability when compared to the voiCe of the person being mimicked. Another purpose of this study was, to determine whether or not there exists identi- fiable similarity when the professional mimic's voice is compared to his own natural voice. .zmmHMOHpuomm venom com mmflhom mo 3mw> unchm .H chamam ill.- _ .- 94:: give I 1.3;... 7,}. pzzuzutt .. Modd 4691 .A.-- ~ Sefles 700 Comparison Spectrograms Figure 2. LOCATION OF MAJOR COMPONENTS (Front) c f _-- ...____. - _. _ ‘ ~" MAGNETIC HEAD I: "1 ASSEMBLY SUPPLY REEL _ 4 ‘. ‘Q fi—TAKE UP REEL A. ; TAPE CLAMP TAPE GUIDES a -' I EXHAUST HOOD 3} TENSIONER I smus BLOCK -——1l SCANNER DRUM STYLUS DRIVE ASSEMBLY MARKING DRUM VU METER MONITOR SPEAKER Figure 3. 4691 Sound Spectrograph Mimicry Biologically, mimicry can have various_functions, one of which is protection. One cannot expect the protection to be absolute, for even genuine warning patterns do not frighten off certain predators. The development of mimetic patterns is often accompanied by other changes, such as be- havior. Psychologically, the human mimics for protection of self-identification and also develops a behavior pattern which accompanies the mimicry. Mimicry serves to deceive the onlooker and listener in the plant and animal kingdom. As a rule, deceptions are successful if they deceive at least once, though the more deceptions the better. Wolf- gang Wickler (1968), in his book on Mimicry in Plants and Animals, states that "one condition of mimicry is the exis- tence of two different signal transmitters...which transmit the same signal and have at least one signal receiver... which reacts similarly to both signal transmitters." One signal transmitter is called the model and the other the mimic. In mimicry of plants and animals one species ex- ploits another, a fact which means that model and mimic must belong to separate species. The human mimics another, either for entertainment or for criminal intent; however, both exploit the person being mimicked. Taxonomy is a science that deals with the classifi- cation of plants and animals according to their presumed natural relationships. Therefore, a taxonomist must decide whether two species or two races are involved in mimicry. In forensic science the qualified examiner in voice identi- fication must decide whether or not two persons are involved in two or more recordings. This examination is accomplished by subjective analysis of sound Spectrograms and aural analysis of the speakers' voices recorded. Jury and judge alike must be convinced beyond a reasonable doubt that the defendant tried to deceive another through mimicry. CHAPTER II REVIEW OF LITERATURE Voice Disguise The most comprehensive study as of this writing in- volving mimicry was performed in 1970 and revised in 1971 by Endres, Bambach and Flosser in Darmstadt, West Germany. This study utilized two well known German mimics who mimicked five speakers including politicians, writers, and actors, all well-known in Germany. The text, 25 seconds in duration, did not correspond word for word among the . five speakers and the text spoken by the mimics. The mimics were allowed to listen to the text spoken by the person to be mimicked and immediately after the mimic was asked to repeat the text, once in mimicry and once in his natural voice. When the original voices were compared with the mimicked voices spectrographically, it was found that the formant structure did not agree, especially in the higher-frequency bands. The mean fundamental frequen- cies also differed. When ten listeners aurally examined these recordings, none was able to tell which was the original voice of the mimic nor could any immediately recognize that the speaker was mimicking another person. The present study, althOUgh quite similar to the Endres, et al. study, used a professional mimic well known 10 11 throughout the United States, Canada and European countries. The mimic's repertoire was used, therefore not restricting what he should say or what vernacular he was to use. The celebrities' conversations were recorded in various settings. Twenty heterogenous college graduates, all of whom knew very well the celebrities and the professional mimic, were utilized as listeners to judge whether or not every pair of voices recorded were spoken by the same person. Automatic Speaker Verification Rekieta and Hair (1972) of Texas Instruments, Inc. and Lummis and Rosenberg (1972) of Bell Laboratories tested several professional mimics using automatic speaker veri- fication methods and found that the mimics were unsuccessful in all their attempts to disguise their voices beyond recog- nition. They used six and eight speakers as control voices, respectively. In the forensic application automatic speaker verification is not a reliable method at this time. The automatic method either accepts or rejects a challenging speaker as a particular one within a short library of voices stored in the memory bank of a computer. The automatic method can be programmed only for closed trials, including a small library of known voices, whereas sound Spectrography used properly is applicable to open trials (voices not limited in number) as well as to closed trials. 12 Appellate Court Reversal On June 25, 1974, the 5th Appellate Court in the State of California in the case of the People vs. E. D. Law re- versed a bomb threat conviction based upon the appellant's estranged wife who testified that the appellant was "very good at imitating many people, many voices...and that the police recording of the telephone bomb threat was not the appellant's 'normal voice', he was imitating someone...I would say, I would say he was trying to imitate a colored person. But he's got a little Oakie in there too." After reviewing the testimonies of Dr's. Oscar Tosi and Peter Ladefoged, the court ruled that when identification is chiefly founded upon an opinion which is derived from _ utilization of an unproven process or technique, the court must be particularly careful to scrutinize the general acceptance of the technique. "A defendant's identity must be established beyond a reasonable doubt...accordingly, the conclusion is compelled that with respect to disguised and mimicked voices in particular the prosecution did not carry its burden of proof to demonstrate that the scientific principles pertaining to spectrographic identification were beyond the experimental and into the demonstrable stage or that the procedure was sufficiently established to have gained general acceptance in the particular scientific field in which it belongs (The Los Angeles Daily Journal Appellate Report, August 1974)." CHAPTER III DESIGN OF STUDY Subjects The professional mimic, Mr. Rich Little, employed in this study is known throughout the United States, Canada and Europe as being the only mimic to have a repertoire of over 150 different men and women. He drew from his reper- toire six celebrities known throughout the motion picture, television and political world. They were Jimmy Stewart, John Wayne, Richard Nixon, Cary Grant, Jack Benny and Johnny Carson, hereafter known as "subjects one through six." They were mimicked by Mr. Little using a similar text with a mean duration of twenty-five seconds. Rich Little, then spoke in his natural voice the same con- textual message (see Appendix B). The recordings from this professional mimic were ac- quired in Ionia, Michigan, in August 1974. Since that time, all of the subjects mimicked were contacted by the writer (see Appendix A); and they were requested to repeat the contextual message mimicked by Mr. Little. Along with a . letter of introduction and instructions they were given the following: a transcript of the mimicked message, a taped excerpt of their mimicked voice, a blank cassette tape for them to record their voice upon and pre-paid envelopes for 13 14 returning the recording. In the last paragraph of the letter of introduction they were told that if for any reason they could not participate, they could return the letter in the pre-paid envelope as acknowledging their non-participation. Most of the subjects refused to record. Mr. John Wayne requested additional instructions (see Appendix A). Therefore, it was necessary to obtain the subjects' voices from pre-recorded interviews, television appearances, and through the assistance of the Michigan State University Voice Library. Dr. Maurice Crane of the Michigan State University Voice Library was contacted; and voice exemplars of most of the subjects, recorded within the year 1974, were acquired from him. The other subjects' voices were acquired by wiring a television set and taping recent appearances of these subjects. A mean duration of thirty- five minutes recorded voice was obtained for each of the subjects mimicked with the exception of Cary Grant. Texts therefore differed, contrary to what it was previously planned. INSTRUMENTATION The recordings of the subjects as well as_the record- ings of the professional mimic were made on cassette magnetic.tapes. A Sony 110A cassette recorder was used for recording the mimic and the recordings obtained from the television. A microphone was used by the mimic, whereas 15 direct wire from the television to the recorder was used to record the subjects. The tapes of the subjects pre- recorded at the Michigan State University Voice Library were copied by an Ampex high-speed duplicator from seven- inch reels of library stored magnetic tape. Once the recordings were made, the recordings were transferred to 1.5 mil polyester magnetic tape on seven- inch reels for spectrographic analysis on the models 700 and 4691 sound speCtrographs. The 300 Hz bandwidth analyzing filter, producing bar-type spectrograms as shown in Figure 2, was used on the 700 series sound spectrograph for the study of the acoustic parameters such as mean frequencies and bandwidths of vowel formants, vertical striations, slopes of formants, duration of similar phonetic elements and plosive gaps, energy distribution of fricatives, plosives and interformant spaces (Figure 4). Fundamental Frequency Within the past 100 years a remarkable series of instruments has been developed by scientists interested in the investigation of the fundamental frequency charac- teristics of vocal and auditory signals. Fundamental frequency is significant, of course, because it is one of the most important components of the complex auditory stimulus that contributes to the perception of the pitch of the sound. The fundamental frequency of the human voice is determined by and is, in fact, identical with 16 Plosive Burst Fricative Fricative FORMANTS 4th 3rd 2nd STOP VOCALCORD VIBRATION PERIOD Figure 4. Display of Acoustic Parameters the rate of vibration of the vocal cords, measured in Hz. The fundamental frequency is the Spectrum's lowest fre- quency component. A fast-acting instrument is required to follow a signal and the instrument must be capable of responding only to the fundamental, or of making a response in which the fundamental is clearly to be distinguished from the other components in the sound. Many ingenious devices, dating back to the mid-nineteenth century, have been designed and constructed for fundamental frequency 17 measurement, as a function of time (Travis, 1971). De- tailed descriptions of all these instruments, while of historical interest, would not contribute much to an under- standing of the instruments and techniques in use today. Although a sizeable number of instruments and methodologies could be reported advantageously within the framework of the communication sciences, the only instrument the writer had available for this study was the Voiceprint Laboratories, Inc., 4691 sound spectrograph with a 45 Hz narrow analyzing bandwidth filter. The fundamental frequency can be easily. determined by utilizing the following graphed formula seen in Figure 5 (Tosi, 1974)., Aural Discrimination An aural discrimination test was given twenty hetero- genous college graduates using a learning services labora- tory classroom at Michigan State University. A Sony 110A cassette recorder was used to play back a pre-constructed tape of twenty randomized paired voices. Their task was to listen to the paired voices and to decide whether or not the paired voices were the same or different. A pre-test questionnaire required each listener to list any hearing impairment that he or she might have. Only one listener stated an impairment: a high frequency hearing loss, which was not significant for the task given. The scores obtained from the aural discrimination task were calcu- lated and an analysis of percentages was tabulated. 18 zucoscohw Hmucoemwcsw mcfiecflw uow «Haasom ecm cameo 3.2:. m2; .uz 93.0. o. 8 Saga. u AH .m ohsmwm IT 000d LT 000” IT 000.? (7H) ADNHOOBHA CHAPTER IV ANALYSIS OF DATA AND CONCLUSIONS Spectrographic Analysis Spectrographic analysis was performed on the natural voice of the professional mimic and the mimicked voices of the six subjects. A total of sixty-nine wide bandwidth spectrograms were produced using the Voice Identification, Inc., series 700 sound spectrograph. Intraspeaker vari- ability was found within the higher frequency bands from approximately 3000 Hz and above. The vowel formants in the lower frequencies were quite similar except for temporal variations because of change in pitch. The slopes of the lower formants were similar with no distinguishable difference within inter-formant structure. The text for each mimicked voice was not verbatim; therefore, the onset and the close of some of the phonetic elements varied be- . cause of the preceding and succeeding sound respectively. Figure 6 shows a spectral display of the mimicked voices of the subjects one through six and left to right, compared to the natural voice of the mimic (spectrum on the far right). I. The natural voices of the six subjects were then com- pared spectrographically with their mimicked voices. A total of eighty-eight wide bandwidth spectrograms were 19 20 ' IIIII'" . III III IIIIIIII I I I “III WILL". I I II , r I III PLAYED PLAYED PLAYED PLAYED PLAYED PLAYED PLAYED Figure 6. Spectra of the six subjects mimicked voices and the professional mimic using his natural voice. made of the subjects' natural voices again using the series 700 sound spectrograph. The interspeaker variability be- tween each subject was found throughout all the spectra, and the interspeaker variability between each subject's natural voice and that of the subject's mimicked voice was as follows: High and low frequency formants were different in bandwidth, slope and duration. The following figures show the differences between each subject's voice and the mimicked voice of that subject. Figures 7 through 10 are the comparisons of subject number five showing the formant differences. Differences in spectra of nasal speech sig- ment is shown in Figures 11 and 12 with subjects five and one respectively. 21 IIIIIIII IIIMI E UH UH Figure 7. Subject number five Figure 8. Mimic and and the mimic uttering the subject number five word /bi/ from the phrase uttering the phoneme F.0.B. and the word /bi/ in //\/ in isolation. isolation, respectively. Figure 9. Mimic and subject Figure 10. Mimic and number five uttering the subject number five word /bAt/. uttering the word /ar/. 22 III: r IAIII‘II’ I * "II I III I I I IIIIIIIIIIIIIIIIII I I .IIIIIII IilIIIIlIIIIIIIIIEIIIIIIIIIIIIII NE V ER NEV ER Figure 11. Subject number five and the mimic uttering the word /n2v€”/. Figure 12. Subject number one and the mimic uttering the word /n£vEr7. 23 Fricative information was found to be as low as 800 - 1000 Hz lower for the mimicked voice than they were for the I natural voice of the subject number three (Figure 13). Vertical striations indicating change of pitch along with voiceless plosives are shown in Figure 14 with the voice of subject number six and the mimicked voice of number six. I WWII III 7 I MUST JOHN JOHN Figure 13. Mimic and subject Figure 14. Mimic and number three uttering the subject number six word /mgst/. uttering word /d)an/. Plasive gaps were wider for the mimicked voice than they were for the natural voice of subject number one (Figure 15). Figures 16 through 21 are again in summary, representative of interspeaker variability between the professional mimic and subjects two, three, four, and six respectively. M .. WW: W % 91 u , 'NH? ” Figure 15. Subject number one and the mimic utter- ing the word /ta:m/. Figure 16. Subject number two and the mimic utter- ing the wdrd /bi/. 25 WW 3 ; h” M : THIS THIS HAVE HAVE Figure 17. Subject number Figure 18. Mimic and sub- one and the mimic utter- ject number three utter- ing the word A?15/. ing the word /ha¢v/. * W Figure 19. Subject number four and the mimic utter- ing the word Aar/. Z6 1|: .,_~____.{ , WWW W ”m NEV ER NEV AND AND Figure 20. Mimic and sub- Figure 21. Subject number ject number six uttering six and the mimic utter- the word /n£v€7. ing the word /acnd/. Fundamental Frequency Analysis The natural voice of the professional mimic, the subjects' natural voices, and the mimicked voices were analyzed on the Voiceprint Laboratories, Inc., 4691 sound spectrograph, using the 45 Hz narrow bandwidth analyzer to determine the speaker's mean fundamental frequency. A total of 103 sound spectrograms were produced. 27 Figure 22 depicts the variance in the mean fundamental frequency between the mimic's natural voice and the mimicked voices of the subjects. 135 130 125 120 115 110 105 100 95 natural voice 1 Z 3 4 S 6 mimicked voices of mimic (99) of subjects. (128,121,132,131,110,123)=fo Figure 22. Mean fundamental frequency between the mimic's natural voice and the mimicked voices of the six subjects. Figure 23 depicts the variance between the mean fundamental frequency of each subject's natural voice and the mean fundamental frequency of each subject's mimicked voice. The natural voice of the subject number four had a mean funda- mental of 212 Hz because the only voice exemplar that was available was a motion picture sound track where the sub- ject was excited and shouting: 28 f0 215 200 185 170 T 155 o o 140 12572—1 110 £i~ 95 r Subjects 1 2 ‘ 3 4 S 6 Natural (144,136,149,212,147,149) Mimicked (128,121,132,131,110,123) ’ Subjects' natural voice 0 Subj ects' mimicked voice a Figure 23. Mean fundamental frequency between the subjects' natural voices and their mimicked voices. Aural Discrimination In addition to the spectrographic analysis an aural test was performed to determine the discriminatory ability of untrained people to discern the difference between the mimicked voice and the natural voice of the subjects. Twenty heterogenous college graduates from Michigan State Univer- sity were chosen and tested in a learning services labora- tory. This laboratory is adequate for grOUp listening in that it is sound isolated and free from ambient reverber- ations. Free field loudness at the ear of each listener was estimated at 70-75 dB. The listeners were given an answer sheet (see Appendix C) with pre-test questions for back- ground information of each listener. The first question was whether or not they had any hearing impairment. Only one 29 student indicated any impairment and that consisted of a high frequency hearing loss which was not significant for the task given. The answer sheet contained twenty line items in which they were to make one of three choices based upon listening to the twenty paired utterances. The choices were same, different, or undecided. The recorded paired voices were excerpts from the re- cordings used for spectrographic analysis and consisted of approximately 10 seconds duration for each utterance within the_pair. One important factor for speaker identification by aural examination is the duration of the speech signal. A speech signal duration of one second or more is sufficient for listener's discrimination (Pollack, Pickett, and Sumby, 1954). The paired recordings did not contain the same con- textual content and were randomized according to a table of random numbers (Dixon and Massey, Jr., 1957). Each mimicked voice by Mr. Little was paired with the voice of the same subject being mimicked or with another mimicked utterance of the same subject. In other words, at no time was the mimicked voice of a given subject paired with the mimicked voice of another subject or the natural voice of a subject paired with the natural voice of another subject. A Sony 110A cassette recorder was used to play the paired recordings, and the test tape was pre-empted with reCorded instructions for the listener as to the answer sheet before them. Before the actual test began, all listeners were questioned whether or not they fully 30 understood their task. Once this was confirmed, the test was then given; and the response sheets were collected immediately thereafter. The paired recordings were classified as being ob- tained with ambient noise or in a quiet atmosphere. Ambient noise as determined for this study includes any recording that contained any other speaker's voice in conjunction with the subject's or the mimic's being recorded. Any music, talking, extraneous noise, audience applause, etc. recorded at the time the speaker was talking was considered as be- longing to the category ambient noise. The signal-to-noise ratio of the ambient noise was estimated to be 6 dB. When a speaker spoke without any ambient noise such recording was classified as quiet. The paired recordings were further classified into three groups called alpha, beta, and gamma. Alpha was the group of paired voices containing the subject paired with the subject; beta was the group of paired voices containing. the mimic paired with the subject; and gamma was the group of paired voices cOntaining the mimic paired with the mimic. Each classification of d , Q , and rcould receive any of the three responses available to the listeners: same, different or undecided. These three different answers were labelled A, B, and C. The following Table 1 depicts the percentage of decisions (A, B, C) made in each general category of d , (I , and r. 31 Table l. Percentages of decisions (A, B, C) as a function of listening for each general category ofol , (9 , and 0’ presentations. 0‘ = subject/subject pairs A = same = right = 79% B = different = wrong = 16% C = undecided = undecided = 5% 0’ = mimic/subject pairs A = same = wrong = 18% B = different = right = 74% C = undecided = undecided = 8% " = mimic/mimic pairs A = same = right = 68% B = different = wrong = 21% C = undecided = undecided = 11% Grand Means: right wrong A undecided 74% 18% 8% Alpha, beta, and gamma classifications were then sub- divided under ambient and quiet recordings as d(,, anduo(‘, Q“ and 9..., and r‘ , and ‘1, respectively. The listeners' answers were collated within these six categories, and it was found that randomization of twenty paired speakers for twenty listeners allows for only one paired voice task in certain categories and as high as seven paired voices in another (Table 2). Table 2. Collation of listeners' answers into ambient and quiet categories. Ambient ' Quiet "I ' ,"'z. 1 paired voice 4 paired voices 20 answers 80 answers 9| 9:. 4 paired voices 7 paired voices 80 answers 140 answers 32 .Table 2. (cont'd.) Y. b". l paired voice 3 paired voices 20 answers 60 answers The percentage of listeners' answers were then cal- culated and the percentages tabulated under eaCh category of ambient noise and quiet (Table 3). Table 3. Percentage of answers (A, B, C) given by listeners in the categories: ambient noise and quiet. Ambient Noise Quiet 0K, 9"- A = 55% A = 85% 04 B = 35% B = 11% C = 10% c = 4% 0; Q. A = 23% A = 15% G B = 70% B = 77% C = 8% C = 8% f: r. A = 30% A = 80% K B = 55% B = 10% C = 15% C = 10% From the preceding tables one can see the significance that ambient noise - quiet situations have upon listener's ability to discriminate between speakers and mimic/speaker. The listeners were able to identify correctly 85% of the paired utterances under quiet conditions as compared to only 55% under ambient noise, when the voices of the same. 33 subject were presented. When the mimicked voice was paired with the subject's voice, 77% correct choices were made under quiet conditions compared to 70% correct choices under ambient noise conditions. When the mimicked voice was paired with the mimicked voice, the listeners were able to choose the correct answer 80% of the time under quiet con- ditions compared to 30% under ambient noise conditions. The listeners began to identify the paired voices correctly as being the same or different after having com- pleted fifty percent of the test, because the mimicked recording contained the same repertoire in context for all six subjects. Only when the paired recordings were dis- similar in acoustic conditions and content did the dis- crimination task become less challenging for the listener. Different acoustical background with each utterance created suspicion against selecting the speakers as being the same. In one paired recording where the mimicked voice of subject six was paired with the mimicked voice of subject six from two separate but similar acoustical backgrounds, the listeners dropped below the grand mean of 74% (Table l) to a low of 30% correct identification. Conclusions The analysis of the data collected reinforces the con- ditions from the previous study by Endres, Bambach, and Flosser and suggests that interspeaker_variability does exist between a mimicked disguised voice and the natural 34 voice of the subject being mimicked. Further, it suggests that the intraspeaker variabilities are minute and not significant when comparing a mimicked voice with the natural voice of the mimic. Further analysis of the aural discrimination results provides the understanding that the smaller the signal-to- noise ratio within the recording and the more similar the contextual material being spoken, the greater the chance for error in distinguishing whether the two speakers are the same or if mimicry is involved. CHAPTER V SUMMARY AND DISCUSSION FOR FUTURE RESEARCH Summar Intraspeaker variabilities of a professional mimic were determined by recording a professional mimic speaking in his natural voice and then mimicking six well known sub- jects. Spectrographic examination was performed to compare the mimic's natural voice with the subjects' mimicked voices. It was found that the acoustic parameters within the spec- trum did not change significantly, suggesting that a pro- fessional mimic could not disguise his voice beyond spectro- graphic recognition. The only intraspeaker variability of significance was found in the higher frequency bands above 3000 Hz. The six subjects were recorded for comparison with the mimicked voice to determine interspeaker variability between each "natural" voice and the same subject's mimicked voice by the professional mimic. The mimicked voices were com- pared spectrographically with the six subjects' natural voices; it was found that there were significant inter- speaker variabilities throughout the spectra, suggesting that it is possible to discriminate spectrographically between a subject's natural voice and his mimicked voice by another person, even a professional mimic. 35 36. The mean fundamental frequency of all subjects' voices, including the natural voice of the mimic and the mimicked voices of the subjects was tabulated. Data showed that though a mimicked voice may acoustically sound the same as the natural voice of the subject, in fact, the pitch is significantly different (Figures 22 and 23). Statistical analysis showed that aural discrimination r1 ability among.the twenty listeners was significant. When - the recordings contained ambient noise or when the text spoken was similar, the listener's ability to discriminate between the paired speakers was impaired. When the record- ings were made under quiet conditions and the texts spoken. were dissimilar, the ability of the listeners to discrim- inate between the paired speakers improved. Conclusions This study further corroborates the findings of earlier studies in Germany on mimicry. Even though automatic methods of identification may be used in the future, a more sub- stantial present day method of identification through sound Spectrography shows validity and reliability through repli- cation. 1 This study demonstrates further that a professional mimic cannot disguise his voite beyond recognition; there- fore, an amateur attempting to disguise his voice through mimicry for any intent, criminal or entertainment, surely can be identified as such by means of sound Spectrography. 37 This finding alone suggests that there is reasonable doubt that a person can disguise his voice through mimicry beyond scientific detection. Recommendations for Future Research Research is being planned for the near future in an extension of this study. An extension of the number of pro- fessional mimics, subjects mimicked, the number of examiners, and an increase in the number of aural discrimination trials given the listeners should be of immediate concern. One important objective is to first acquire the re- corded voices of the subjects to be mimicked. The major ‘ ’ constraint of this study was the lack of cooperation of the celebrities because funds were not available. The busy itineraries of most professionals are such that proper fund- ing of any future study must be considered. Pre-recorded subjects' voices are more economically available and are sufficient, providing the category of quiet conditions be- comes mandatory. Once the subjects' voices have been recOrded, the pro- fessional mimics should be considered. Since Mr. Little cooperated, mimics such as Frank Gorshin, David Frye, and George Kirby may respond less reluctantly. Again, proper funds must be considered along with direct contact with the mimics themselves. Acquiring trained examiners from the International Association of Voice Identification to take part in the '38 spectrographic and aural analysis of the mimicked and natural voices should be considered. A cross-represenv tation of examiners throughout the United States would add credence to the validity of this study. Aural examinations by listeners provided with mone- tary incentive should also be considered. Utilizing an anechoic chamber or listening cells equipped with headsets F} for discrete discrimination would also be preferred, rather than group listening as in this study. A time constraint existed in this study, whereas in the future this constraint could be eliminated by monetary remuneration. Extended numbers of randomized paired speakers providing more trials in each category of alpha, beta, and gamma would give better statistical data. APPENDICES APPENDIX A CORRESPONDENCE WITH MIMIC AND CELEBRITIES 39 APPENDIX A CORRESPONDENCE WITH MIMIC AND CELEBRITIES STATE OF MICHIGAN (“W T) ‘17 42:?) Ir WILLIAM G. MILLIKEN. GOVERNOR DEPARTMENT OF STATE POLICE . 714 s. HARRISON 90.. EAST LANSING. MICHIGAN 48823 COL. JOHN R. PLANTS. DIRECTOR May 16, 1974 Creative Management '600 Madison Avenue New York, New York 10019 Gentlemen: The Voice Identification Unit of our Scientific Crime Laboratory Section is doing further research on the identification of voices. Det, Sgt. Malcolm Hall of this Unit is using this research material for his Masters thesis. I have been informed that your agency represents Mr. Rich Little--the television and movie celeb- rity. Would Mr. Little make a recording for us on magnetic tape of any conversation using his normal voice. Then, repeat the same conversation in the voice of a number of celebrities he impersonates. The length of each sample and the number of samples would be left to the discretion of Mr. Little. Since funding for this research is limited, and is the responsibility of D/Sergeant Hall, please advise the cost that would be incurred prior to rendering this service. Sincerel /( @232; /’ DIRECTOR 0/ 40 STATE OF MICHIGAN r”)! I ‘4 l ("‘4‘ WILLIAM G. MILLIKEN. GOVERNOR DEPARTMENT OF STATE POLICE 714 S. HARRISON RD.. EAST LANSING. MICHIGAN 48823 COL. JOHN R. PLANTS. DIRECTOR ' I‘ May 16, 1974 William Morris Agency ' f Authorization Department L1 1350 Ave. of the Americas New York, New York 10019 Gentlemen: The Voice Identification Unit of our Scientific Crime Laboratory Section is doing further research on the identification of voices. Det. Sgt. Malcolm Hall of this Unit is using this research material for his Masters thesis. I have been informed that your agency represents Mr. Frank Gorshen--the television and movie celeb- rity. Would Mr. Gorshen make a recording for us on magnetic tape of any conversation using his normal voice. Then, repeat the same conversation in the voice of a number of celebrities he impersonates. The length of each sample and the number of samples would be left to the discretion of Mr. Gorshen. Since funding for this research is limited, and is the responsibility of D/Sergeant Hall, please advise the cost that would PI incurred prior to rendering this service. Sincerel , , ,{W/ DIRECTOR K 41 BNB ASSOCIATES LTD. NINETY FOUR FIFTY FOUR WILSHIRE BLVD. BEVERLY HILLS. CALIFORNIA 90212 (213) 273-7020 CABLE: SHERMACE May 29, 1974 Mr. John Plants Director DEPARTMENT OF STATE POLICE 714 South Harrington Road East Lansing, Michigan 48823 Dear Mr. Plants: In answer to your letter of Mby 16, 1974 which was forwarded to us from the William Morris Agency regarding Frank Gorshin. We are Mr. Gorshin's personal managers, and we are very sorry to infbrm you that Mr. Gorshin is tied up for the next several months making personal appearances both here and in Hawaii, plus some television shows. Unfortunately this schedule does not allow him enough time to comply with your request. Mr. Gorshin is very flattered and intriuged by your inquiry and appreciates your interest in him. Best regards. Si ere/1y. , Qty/WU ' 14W Mar ret Prestel 42 August 8, 1974 Mr. Rich Littlev 9000 Sunset Blvd. Los Angeles, California 90069 Dear Mr. Little: Enclosed are six original letters for your signature. These are to be sent to the following people, along with a recording of your mimic of them; which you so graciously provided for me in Ionia, Michigan. Jimmy Stewart Jack Benny John Wayne Johnny Carson Cary Grant . President Nixon If you could provide the addresses of the above, with the exception of President Nixon, I would be greatly indebted to you. In addition to your letter, the above celebrities will receive a letter from the Director of the Michigan State Police, and a set of instructions from myself. If my intentions and/or the format of the letter submitted for your signature does not meet with your approval, please do not hesitate to make the desired changes, or advise me of your preference. Again, I thank you for your cooperation and the interest you have shown, regarding my thesis and experi- mental endeavor. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEII: taw Encl. ‘_. In” lpwm‘dlgérfl. a 43 Dear Det. Sgt. Malcolm Hall, of the Michigan State Police —T Voice Identification Unit, and I recently corroborated on a recording involving professional mimicry. Your voice was one of six well known celebrities selected. I believe this present study, involving comparison and identification, will be beneficial to this relatively new science and progressive law enforcement today. I respect- fully request your cooperation with Detective Sergeant Hall in his experimental endeavor. Sincerely, 44 August 12, 1974 . Mr. Rich Little 9000 Sunset Blvd. Los Angeles, California 90069 F Dear Mr. Little: Enclosed are the spectrograms of your voice which I promised earlier. I'm sorry I forgot to send them with the small gift of appreciation sent recently. - Also enclosed, please find a return addressed, postage paid envelope for the return of the letters previously sent. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Encl. 45 November 7, 1974 Mr. Mel Bishop Sands Hotel Las Vegas, Nevada Dear Mr. Bishop: Reference our telephone conversation this date, please find enclosed the six (6) original letters for Mr. Little's signature. It would be greatly appreciated if your person- nel in California could provide me with the addresses of the following people, or the addresses of their agents: Jimmy Stewart Jack Benny John Wayne Johnny Carson Cary Grant Richard M. Nixon As previously mentioned, I will construct a cover letter which will accompany your letter, along with in- structions for the project, to each celebrity. It is imperative that this portion of the study be completed as soon as possible, so that the thesis can be completed by the end of this year. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Encl. I" 46 December 12, 1974 James Stewart 918 North Roxbury Drive Beverly Hills, California Dear Mr. Stewart: By way of introduction, my name is Malcolm Hall. I am a graduate student at Michigan State University and currently employed as a detective sergeant with the Michigan Depart- ment of State Police, as a forensic laboratory scientist. This letter is in reference to a collaborative study of professional mimicry with Rich Little. Mr. Little recently recorded for me, mimicked voices of several celebrities of which yours was one; therefore, I am requesting your parti- cipation in this study. I must add, there are no funds available from either institution for your remuneration; I can only hope for your cooperation, as I did Mr. Little's, for the advancement of a relatively new scientific identi- fication process within the criminal justice system. I am requesting that you take a few minutes of your time to listen to the enclosed cassette recording marked #1, which contains Rich Little's mimic of your voice. Follow the en- closed written transcript, as you listen, so that you can better understand the vernacular he used. 'Second, using the new cassette tape marked #2, record the mimicked message verbatim at least three times (holding the microphone two- to- three inches from your mouth). It is important that you make the recording as soon as pos- sible, so that the voices compared in this study will re- main contemporary. Furthermore, I have a time limit on completing my thesis requirements, which is nearing an end. If for any reason you cannot participate in this study, please return the enclosed letter of request to acknowledge your receipt of it. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Enc. I III I i I lllllllll'lllll‘. I 47 December 12, 1974 John Wayne 2682 Bayshore Drive Newport Beach, California Dear Mr. Wayne: By way of introduction, my name is Malcolm Hall. I am a graduate student at Michigan State University and currently employed as a detective sergeant with the Michigan Depart- ment of State Police, as a forensic laboratory scientist. ' This letter is in reference to a collaborative study of professional mimicry with Rich Little. Mr. Little recently recorded for me, mimicked voices of several celebrities of which yours was one; therefore, I am requesting your parti- " cipation in this study. I must add, there are no funds | available from either institution for your remuneration; I fifi can only hope for your cooperation, as I did Mr. Little's, for the advancement of a relatively new scientific identi— fication process within the criminal justice system. I am requesting that you take a few minutes of your time to listen to the enclosed cassette recording marked #1, which contains Rich Little's mimic of your voice. Follow the en- closed written transcript, as you listen, so that you can better understand the vernacular he used. Second, using the new cassette tape marked #2, record the mimicked message verbatim at least three times (holding the microphone two- to-three inches from your mouth). It is important that you make the recording as soon as pos- sible, so that the voices compared in this study will re- main contemporary. Furthermore, I have a time limit on completing my thesis requirements, which is nearing an end. If for any reason you cannot participate in this study, please return the enclosed letter of request to acknowledge your receipt of it. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Enc. 48 December 12, 1974 Richard Nixon San Clemente, California Dear Mr. Nixon: By way of introduction, my name is Malcolm Hall. I am a graduate student at Michigan State University and currently employed aspa detective sergeant with the Michigan Depart- ment of State Police, as a forensic laboratory scientist. FE This letter is in reference to a collaborative study of professional mimicry with Rich Little. Mr. Little recently recorded for me, mimicked voices of several celebrities of which yours was one; therefore, I am requesting your parti— cipation in this study. I must add, there are no funds available from either institution for your remuneration; I can only hope for your cooperation, as I did Mr. Little's, E for the advancement of a relatively new scientific identi- fication process within the criminal justice system. I am requesting that you take a few minutes of your time to listen to the enclosed cassette recording marked #1, which contains Rich Little's mimic of your voice. Follow the en- closed written transcript, as you listen, so that you can better understand the vernacular he used. Second, using the new cassette tape marked #2, record the mimicked message verbatim at least three times (holding the microphone two- to-three inches from your mouth). It is important that you make the recording as soon as pos- sible, so that the voices compared in this study will re- main contemporary. Furthermore, I have a time limit on completing my thesis requirements, which is nearing an end. If for any reason you cannot participate in this study, please return the enclosed letter of request to acknowledge your receipt of it. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Enc. Illl‘Il‘I ll 49 December 12, 1974 Cary Grant P. O. Box 1551 Beverly Hills, California Dear Mr. Grant: By way of introduction, my name is Malcolm Hall. I am a graduate student at Michigan State University and currently employed as a detective sergeant with the Michigan Depart- N‘ ment.of State Police, as a forensic laboratory scientist. This letter is in reference to a collaborative study of professional mimicry with Rich Little. Mr. Little recently recorded for me, mimicked voices of several celebrities of which yours was one; therefore, I am requesting your parti- cipation in this study. I must add, there are no funds available from either institution for your remuneration; I can only hope for your cooperation, as I did Mr. Little's, for the advancement of a relatively new scientific identi- fication process within the criminal justice system. I am requesting that you take a few minutes of your time to listen to the enclosed cassette recording marked #1, which contains Rich Little's mimic of your voice. Follow the en- closed written transcript, as you listen, so that you can better understand the vernacular he used. Second, using the new cassette tape marked #2, record the mimicked message verbatim at least three times (holding the microphone two- to-three inches from your mouth). It is important that you make the recording as soon as pos- sible, so that the voices compared in this study will re- main contemporary. Furthermore, I have a time limit on completing my thesis requirements, which is nearing an end. If for any reason-you cannot participate in this study, please return the enclosed letter of request to acknowledge your receipt of it. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Enc. 50 December 12, 1974 Jack Benny 10231 Charing Cross Road Beverly Hills, California Dear Mr. Benny: By way of introduction, my name is Malcolm Hall. I am a graduate student at Michigan State University and currently employed as a detective sergeant with the Michigan Depart- 1 ment of State Police, as a forensic laboratory scientist. This letter is in reference to a collaborative study of professional mimicry with Rich Little. Mr. Little recently recorded for me, mimicked voices of several celebrities of which yours was one; therefore, I am requesting your parti- cipation in this study. I must add, there are no funds available from either institution for your remuneration; I can only hope for your cooperation, as I did Mr. Little's, for the advancement of a relatively new scientific identi- fication process within the criminal justice system. I am requesting that you take a few minutes of your time to listen to the enclosed cassette recording marked #1, which contains Rich Little's mimic of your voice. Follow the en- closed written transcript, as you listen, so that you can better understand the vernacular he used. Second, using the new cassette tape marked #2, record the mimicked message verbatim at least three times (holding the microphone two- to-three inches from your mouth). It is important that you make the recording as soon as pos- sible, so that the voices compared in this study will re- main contemporary. Furthermore, I have a time limit on completing my thesis requirements, which is nearing an end. If for any reason you cannot participate in this study, please return the enclosed letter of request to acknowledge your receipt of it. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Enc. 51 December 12, 1974 Johnny Carson .‘ 400 St. Cloud Road Belair, California Dear Mr. Carson: By way of introduction, my name is Malcolm Hall; I am a graduate student at Michigan State University and currently employed as a detective sergeant with the Michigan Depart- ment of State Police, as a forensic laboratory scientist. 1..., This letter is in reference to a collaborative study of professional mimicry with Rich Little. Mr. Little recently recorded for me, mimicked voices of several celebrities of which yours was one; therefore, I am requesting your parti- cipation in this study. I must add, there are no funds -' i available from either institution for your remuneration; I can only hope for your cooperation, as I did Mr. Little's, for the advancement of a relatively new Scientific identi- fication process within the criminal justice system. I am requesting that you take a few minutes of your time to listen to the enclosed cassette recording marked #1, which contains Rich Little's mimic of your voice. Follow the en- closed written transcript, as you listen, so that you can better understand the vernacular he used. Second, using the new cassette tape marked #2, record the mimicked message verbatim at least three times (holding the microphone two- to-three inches from your mouth). It is important that you make the recording as soon as pos- sible, so that the voices compared in this study will re- main contemporary. Furthermore, I have a time limit on completing my thesis requirements, which is nearing an end. If for any reason you cannot participate in this study, please return the enclosed letter of request to acknowledge your receipt of it. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Enc. is: 52 January 7, 1975 Mr. Richard Nixon San Clemente, California Dear Mr. Nixon: On December 16, 1974, you received a package contain- ing two cassette tapes with a request that you record a 1 short narrative which had been mimicked by Rich Little. 7 I would like to thank you for taking time from your holiday season to accept the package, and eagerly await the return of your tape to collate the results. Further, h] it is my pleasure to inform you that Rich Little is to co-host the "Tonight Show" this month and is considering introducing the results of this study. Mr. Little and his manager, Mel Bishop, are both very interested in this study and are eager to see it completed as soon as possible. I would like to thank you again for accepting the package, and ask that you return the tape as soon as pos- sible as time is an important factor in this study. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw 53 January 7, 1975 Mr. Johnny Carson 400 St. Cloud Road Belair, California Dear Mr. Carson: On December 17, 1974, you received a package contain- ing two cassette tapes with a request that you record a short narrative which had been mimicked by Rich Little. I would like to thank you for taking time from your holiday season to accept the package, and eagerly await the return of your tape to collate the results. Further, it is my understanding that Rich Little is to co-host the "Tonight Show" this month and is considering introducing the results of this study. Mr. Little and his manager, Mel Bishop, are both very interested in this study and are eager to see it completed as soon as possible. I would like to thank you again for accepting the package, and ask that you return the tape as soon as pos- sible as time is an important factor in this study. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw rr ”211.51- I. 54 JOHN WAYNE 9570 Wilshire Blvd. , Suite 400 Beverly Hills, California 90212 January 3, 1975 Mr. Malcolm E. Hall 714 S. Harrison Rd. East Lansing, Michigan 48823 Dear Mr. Hall: Mimicry is a form of flattery, and I feel very complimented every time that Rich Little mimics me for the entertainment of others. However, if you wish to prove that it can be another snag in the procedure for justice, I suggest that you send me a phrase which you would have me read rather than have me attempt to mimic Rich Little mimicking me. Because of that reasoning, I have not listened to the enclosed tape recording. Sinc er ely,“ JW/ps 55 January 9, 1975 Mr. John Wayne 2682 Bashore Drive Newport Beach, California 92660 Dear Mr. Wayne: In reference to your letter dated January 3, 1975, I would like to clarify my instructions. I did not intend for you to remimic Rich Little, only repeat in your normal voice his mimic of you. The transcript was sent for your assis- tance to insure keeping the context verbatim. The tape was for your listening pleasure and to better understand the vernacular used. L Please accept these tapes, one of which is to be used for your recording, and transcript of which I would like you to read from in your normal.voice.A I'm sorry the instructions were not explicit enough, and would like to thank you again for your patience and interest thus far. Sincerely, Malcolm E. Hall Detective Sergeant Michigan State Police Voice Identification Unit MEH:taw Enc. APPENDIX B RAW DATA OF MIMICKED TEXT 56 APPENDIX B RAW DATA OF MIMICKED TEXT AH THIS, AH THIS IS JIMMY JIMMY STEWART. I-EH IT, IT CERTAINLY IS IS GREAT THRILL DA BE HERE IN IONIA. AN I I'VE, I'VE NEVER BEEN IN THIS PART OF THA COUNTRY BEFORE. I PLAYED DETROIT AND AN CHICAGO MANY TIME BUT I, I'VE JUST I'VE JUST NEVER BEEN ABLE TA GET UP HERE. AN, AN I MUST SAY IT'S, IT IT'S BEAUTIFUL COUNTRY. WELL THIS IS JOHN WAYNE. IT SURE IS NICE TUH BE HERE IN IONIA. I'VE NEVER PLAYED THIS PART OF THA COUNTRY BAFORE. I BEEN DA CHICAGO AN --DETROIT MANY TIMES BUT I GOTTA TELL --YA IT SURE IS MIGHTY PRETTY COUNTRY. WHEN I CAME OUT HERE TO TEXAS I WAS LOOKIN FOR SUMP' N MISTER I'M JOHN WAYNE AN THIS IS--AN THE NEXT GUY WHO HONKS I'M GONA SHOOT 'EM RIGHT BETWEEN ... UH, THIS IS RICHARD NIXON I HAVE NEVER EH PLAYED THIS UH THIS PART UH-THA COUNTRY BE FORE AN THA WAY THINGS ARE GOING I DON'T THINK I EVER WILL. BUT I MUST SAY THIS, IT'S UH VERY BEAUTIFUL COMMUNITY MAKE NO MISTAKE ABOUT THAT. UH, LET'S SEE WHO ELSE-~CARY GRANT. AH YES, YEAH THIS IS THA FIRST TIME I'VE EVER BEEN IN IONIA. I'VE NEVER BEEN HEREBEFORE. I PLAYED DETROIT AN CHICAGO MANY TIMES. I NEVER I NEVER BEEN ABLE TA GET UP HERE AND I MUST SAY IT'S BEAUTIFUL COUNTRY. 57 JACK BENNY. GOSH YA-KNOW, IT REALLY IS UH BIG THRILL DA BE HERE IN IONIA. EH I'M PERFORMING HERE FOR NOTHING. BUT UH I PLAYED DETROIT AN CHICAGO, BUT I'VE NEVER--- BEEN DA THIS PART UH THA COUNTRY. AND IT'S BEAUTIFUL. AH, THIS IS JOHNNY CARSON I AND DUH = I'VE NEVER UH NEVER BEEN TO IONIA BEFORE IT'S IT'S UH BEAUTIFUL LITTLE TOWN IF THEY IF THEY EVER GET IT FINISHED. AH (LAUGHING) NO, I PLAYED DETROIT AN CHICAGO BUT NEVER BEEN HERE. 1, IT WAS 50 HOT HOW HOT WAS IT? JOHN-JOHN- JOHNNY CARSON MAY UH CAMEL WITH UH WEAK KIDNEY BACK UP INTO YOUR BREAKFAST I WOULD SUGGEST THAT YOU WATCH WIDE WIDE WORLD OF ENTERTAINMENT- IN - INNUH SEPARATE ROOM. LUCY IS ADORABLE, SHE'S BEAUTIFUL, SHE'S WONDERFUL AND SHE'S FABULOUS, AND DUH, I ONLY WISH HER NAME WAS JOANNA, THEN I WOULD OF MARRIED HER. APPENDIX C RAW DATA OF AURAL DISCRIMINATION TEST' 58 AURAL DISCRIMINATION TEST YOUR NAME AGE MALE OR FEMALE HEARING IMPAIRMENT IF ANY PAIRED VOICES SAME DIFFERENT UNDECIDED OF SPEAKER(S) » 1. 2. 3. 10. ll. 12. 13. 14. 15. 16. 17. 18. 19. 20. LIST OF REFERENCES~ 59 LIST OF REFERENCES Daniloff, R. G. and Hammarberg, R. B., ”On Defining Co- articulation", Journal of Phonetics, I, pp. 185-194 (April 1973). Dixon, W. J. and Massey, Jr., F.J., Introduction to Statistical Analysis. McGraw-Hill Book Co., Inc. New York.' Appendix 513, (1957). Endres, W., Bambach, W., and Flosser, 6., "Voice Spectro- grams as a Function of Age, VOice Disguise, and Voice Imitation", Journal of the Acoustical Society of America, 49, pp. 1842-1848 (1971). Kamine, B. 8., "The Voiceprint Technique: Its Structure and Reliability", San Diego Law Review, 6, 2, pp. 213-241 (May 1969). Los Angeles Daily Journal, The, "Appellate Report", Tuesday, August 13, 1974. In the Court of Appeal Of the State of California Fifth Appellate District. Lummis, R. C. and Rosenberg, A. B., "Test of an Automatic Speaker Verification Method with Intensively Trained Professional Mimics". Journal of the Acoustical Society of America, 51, p. 131 (1972). Pollack, I. Pickett, J., and Sumby, W., "On the Identi- fication of Speakers by Voice", Journal of the Acoustical Societygof America, 26, pp. 403-406 (1954). Presti, A. J., "High Speed Sound Spectrograph",Journal of the Acoustical Society of America, 40, pp. 628-634 (1966). Rekieta, T. W., and Hair, G. D., "Mimic Resistance of Speaker Verification Using Phoneme Spectra", Journal of the Acoustical Society of America, 51, p. 131 i1972). Tosi, O., Oyer, H., Lashbrook, W., Pedrey, C., Nicol, J., and Nash, E., "Experiment on Voice Identification", Journal of the Acoustical Society of America, 51, pp. 2030-2043, (1972). Tosi, 0., "Longitudinal Study of Childrens Voices at Puberty", Postan D. Bianculli J., Paper presented at XVI Inter- national Congress of Logopedics and Phonatrics, Interlochen, Switzerland, August, 1974. III I I I. I III} 1 II. II III A. . II I! i [III] III ‘ h I I-IA:.I:PFIIR'~ 1.1!». ’N 60 Travis, Edward Lee., Handbook of Speech Pathology and Audiology. Appleton-Century-Crofts, Meridith Corp., New York, (1971). Wickler, Wolfgang, Mimicry in Plants and Animals, translated from German by R. D. Martin, World University Library, McGraw-Hill Book Co., New York-Toronto, pp. 235-243, (1968). ' ll li i l