NCJ Number
64619
Journal
FBI Law Enforcement Bulletin Volume: 49 Issue: 1 Dated: (JANUARY 1980) Pages: 1-4
Date Published
1980
Length
4 pages
Annotation
SPEAKER IDENTIFICATION METHODS FALL INTO THREE GROUPS--A LISTENING PROCESS, MACHINE ANALYSIS, AND AURAL-VISUAL COMPARISON USING SPEECH SPECTROGRAMS; EACH METHOD HAS DRAWBACKS AND ADVANTAGES.
Abstract
FROM THE POLICE STANDPOINT, IDENTIFICATION BY LISTENING HAS NUMEROUS DRAWBACKS. THE TECHNIQUE CAN BE VERY SUBJECTIVE AND IS NOT COMPLETELY UNDERSTOOD. IT IS OFTEN DIFFICULT FOR WITNESSES TO EXPLAIN THE BASIS FOR THEIR IDENTIFICATION TO A JURY. HOWEVER, IF WITNESSES CAN EXPRESS A BASIS FOR IDENTIFYING THE SPEAKER'S VOICE, TESTIMONY IS ADMISSIBLE IN MOST COURTS. THE WEIGHT GIVEN THE TESTIMONY BY JUDGE AND JURY DEPENDS ON A NUMBER OF FACTORS INCLUDING THE WITNESS' FAMILIARITY WITH THE DEFENDANT'S VOICE. MACHINE ANALYSIS AND COMPARISON, ON THE OTHER HAND, COULD BECOME A MORE OBJECTIVE AND ACCURATE METHOD FOR SPEAKER IDENTIFICATION. IT IS CURRENTLY USED FOR VERIFICATION, RATHER THAN IDENTIFICATION, AS IN BANK-BY-PHONE SYSTEMS. A THIRD METHOD OF VOICE IDENTIFICATION USES SPEECH SPECTROGRAMS OR VOICEPRINTS TO MAKE AN AURAL-VISUAL COMPARISON OF TWO VOICES. WHEN CONDUCTING SUCH A COMPARISON, THE EXAMINER USES THE SOUND SPECTROGRAPH TO PRODUCE SPECTROGRAPHS OF THE UNKNOWN AND KNOWN VOICES SAYING THE SAME WORDS AND PHRASES. THE SPECTROGRAMS ARE THEN MARKED BY WRITING EACH SYLLABLE AND WORD IN THE SPEECH DIRECTLY BELOW THE CORRESPONDING PATTERNS ON THE SPECTROGRAMS. PARTICULAR SPEECH SEGMENTS ARE THEN COMPARED FOR SIMILARITIES AND DISSIMILARITIES IN SPECTROGRAPHIC PATTERNS AND VOICE CHARACTERISTICS. COMPARISONS TO DETERMINE IF SEVERAL UNKNOWN VOICES BELONG TO THE SAME PERSON MAY BE MADE. FEDERAL AND STATE APPELLATE COURT DECISIONS ARE DIVIDED ON THE ACCEPTABILITY OF VOICE SPECTROGRAMS AS COURT EVIDENCE, A DIVISION THAT IS BASED ON THE AMBIGUITY OF LEGAL RULES CONCERNING THE INTRODUCTION OF NEW SCIENTIFIC EVIDENCE IN COURT. FOOTNOTES ARE GIVEN. (AOP)