JANUARY 1980 ---,. ­ LAW ENFORCEMENT BUllETIN JANUARY 1980, VOLUME 49, NUMBER 1 Contents Forensic Science 1 Speaker Identification (Part 1) Three Methods- Listening, Machine, and Aural­Visual By Bruce E. Koenig, Special Agent, Technical Services Division, Federal Bureau of Investigation, Washington, D.C. 8 Hypnosis: The FBI's Team Approach By Richard L. Ault, Jr., Special Agent, Behavioral Science Unit, FBI Academy, Quantico, Va. Crime Problems 9 Automobile Theft: An Increasing Crime Problem By Samuel J. Rozzi, Commissioner of Police, Nassau County, N.Y. , and Det. Sgt. Richard Mueller, Police Department, Nassau County, N.Y. Facilities 14 The Canadian Police College By Charles W. Steinmetz, Special Agent, Education and Communication Arts Unit, FBI Academy, Quantico, Va. Point of View 19 Higher Education for Police Officers By Thomas A. Reppetto, Ph. D., Vice President and Professor of Criminal Justice Administration, John Jay College, New York, N.Y. The Legal Digest 28 The Constitutionality of Routine License Check Stops- A Review of Delaware v. Prouse By Daniel L. Schofield, Special Agent, Legal Counsel Division, Federal Bureau of Investigation, Washington, D.C. 2S Wanted by the FBI The Cover: Federal Bureau of Investigation Published by the Public Affairs Office, Voiceprints­finger- United States Department of Justice Homer A. Boynton, Jr., prints of the future or Executive Assistant Director Washington, D.C. 20535 investigative tool for Editor-Thomas J. Deakin today? See story William H. Webster, Director Assistant Editor-Kathryn E. Sulewski page 1. Art Director-Carl A. Gnam, Jr. Writer/Editor-Karen McCarron The Attorney General has determined that the publication Production Manager-Jeffery L. Summers of this periodical is necessary in the transaction of the public business required by law of the Department of Justice. Use of funds for printing this periodical has been approved by the Director of the Office of Management and Budget through December 28, 1983. ISSN 0014­5688 USPS 383­310 SPEAKER IDENTIFICATION (Part 1) The 20th century has witnessed alarms, and bomb threats are only a the rapid development and prolifera­ few ways in which criminals can perpe­ Three tion of communications technology­ trate their acts through use of the tele­ radio, telephone, television, recording phone. Often, the only physical Methods­ devices-which have profoundly influ­ evidence linking an individual or group enced the manner in which individuals to a crime is a recording of one or more Listening, communicate. The telephone system voices by the police department or of modern, urban society permits in­ victim. Machine, and stantaneous communication between If the police later develop sus­ virtually every business and household pects in the investigation, they are in the community and is especially in­ faced with the prospect of determining Aural- Visual fluential in the way we pass on infor­ culpability based on comparisons of mation. Therefore, it is not surprising the voices. For this reason, lawen­ that of the various communications forcement officials have an increasing media, the telephone, in particular, is need for objective, reliable methods of BY BRUCE E. KOENIG adopted by criminals, both as an aid to identifying telephone callers and the subsequent information which can be Special Agent committing their crimes and as a means of maintaining anonymity. used for investigative leads and court­ Technical SeNices Division room purposes. Consequently, devel­ Federal Bureau of Investigation Telephoning instructions for a ran­ opments in forensic applications of Washington, D.C som payoff in a kidnaping case, trans­ mitting wagering information, speaker identification are of consider­ telephoning the police or news media able interest to the law enforcement to claim credit for a terrorist action, and community. communicating extortionate or ob­ scene telephone calls, false fire January 1980 / 1 A sound spectrograph gives a visual display of human speech. Speaker identification, as used in or to human interaction which may be From the standpoint of usefulness this article, refers to a broad realm of employed in the decision process. to the police, identification by listening processes, both human and electronic, Identifying a speaker by listening has numerous drawbacks. The tech­ by which voices are analyzed, com­ is familiar to everyone because that is nique can be very subjective, the proc­ pared, and decisions of identity made. the method used on a daily basis. Iden­ ess by which people identify voices is Speaker identification methods tification by listening might occur in not completely understood, and it is have been more or less grouped into one of many forms. For example, the often difficult for the witness to convey three general categories-listening, recipient of obscene or harassing tele­ adequately the basis for his identifica­ machine analysis, and aural-visual phone calls may be called on to listen tion to a jury. Some speech scientists comparison using speech spectro­ to the voices of several suspects (a believe, however, that identification by grams. The first and third methods are "voice lineup"). Friends, relatives, or listening may be an accurate and reli­ essentially human processes, depend­ business associates of a suspect may able voice identification method in cer­ ing on the sensory faculties of hearing listen to a voice recording of an un­ tain circumstances.! Testimony by a and sight and the comparative abilities known caller to determine whether witness that he could identify a speak­ of the human brain. The second meth­ they could identify the unknown voice er's voice has been admissible in most od employs electronic devices to ana­ as being that of the suspect. Police courts in this country since at least the lyze the human voice, extract certain officers, listening to recordings made early 1900's, as long as the witness features that may be associated with during a court-authorized wiretap in a was able to express the basis for iden­ the identity of the speaker, and com­ gambling or narcotics investigation, tity.2 The weight given the testimony by pare these features against similar fea­ may identify one of the callers as being judge and jury depends on a number of tures extracted from other voice a subject whom they had extenSively considerations, including the witness' samples. The decision process may be interviewed in a previous investigation. familiarity with the defendant's voice. left either entirely to the machine, in Machine analysis and comparison which case the success or failure of has the potential for being a more the technique may depend on how objective and accurate method for cleverly the machine was programed, speaker identification.3 However, for 2 I FBI Law Enforcement Bulletin Figure 1 Examiner compares spectrograms for similarities. the present, it seems to be a more special paper and is coupled to a mag­ band. As the drum revolves, the vari­ useful technique for verification of a netic tape playback device, and (4) an able electronic filter passes increasing­ cooperative speaker's voice when electric stylus which marks the paper ly higher frequencies with the electric compared to a relatively small group of as the drum rotates. The magnetic stylus moving parallel to the axis of the known voice samples. This type of tape recorder is first used to record the drum. (See fig. 1.) Thus, a pattern of speaker "verification" rather than speech sample onto a reel of magnetic closely spaced lines is generated on "identification" has a number of com­ tape. Next, a short loop of the record­ the paper. The pattern is usually re­ mercial applications, for example, con­ ed tape containing the portion of ferred to as a spectrogram and has the firmation of identity in bank-by-phone speech of interest is wrapped around dimensions of frequency, time, and systems. Consequently, for the most the playback device on the drum, and amplitude. 5 part, research in this area has been the speech sample on the drum is then The vertical dimension in figure 2 directed at achieving an accurate verifi­ played repeatedly at high speed in or­ shows frequency from 0-8000 Hertz or cation process. der to analyze its spectral or frequency cycles per second, the horizontal di­ content. For each revolution of the mension shows time with each of the Voiceprint Identification drum, the variable electronic filter short vertical bars near the top desig­ The third method of voice identifi­ passes only a certain band of frequen­ nating 0.1 second, and the darkness of cation uses speech spectrograms, or cies, and the energy in this frequency the trace indicates the energy of the what are commonly called "voice­ band activates the electric stylus so voice at that time and frequency. prints," 4 to make an aural-visual com­ that a spiral of varying darkness is When conducting an aural-visual parison of two voices. The tool of the produced across the paper. The dark­ voice identification of two or more voiceprint technique is a sound spec­ ness of the line at any point on the voices, the VOiceprint examiner uses trograph which gives a visual display of paper indicates the amount of energy the sound spectrograph to produce the complex waveforms in human present in the speech signal at the spectrograms of the unknown and speech. The sound spectrograph con­ specific time within the given frequency known voices saying the same words sists of four basic parts: (1) A magnetic and phrases. The spectrograms are tape recorder, (2) a variable electronic then marked by writing each syllable filter, (3) a drum which holds a sheet of and word in the speech directly below January 1980 / 3 I f Figure 2 the corresponding patterns on the The above procedure requires that the acceptability of voice spectrograms spectrograms. the examiner look for unusual patterns as evidence in court. This acceptability The voiceprint examiner then uses that may be important clues for identifi­ problem seems to center on the ambi­ one of several methods to conduct the cation.