NCJ Number
57710
Date Published
1976
Length
7 pages
Annotation
THE BASIS FOR USING VOICEPRINT TO IDENTIFY SPEAKERS IS DISCUSSED, AS WELL AS AN EXPERIMENT WITH 25 ADULT MALE SPEAKERS CONDUCTED TO ESTABLISH THE EFFECTIVENESS OF VOICEPRINT IN DETECTING SPEAKER DISGUISE.
Abstract
WHEN SPEAKING, AN INDIVIDUAL PRODUCES A COMPLEX ACOUSTIC SIGNAL CONTAINING INFORMATION ON THE PERSON'S ANATOMY, THE PHYSIOLOGY OF THE SPEAKER'S LARYNGEAL SOURCE AND VOCAL TRACT, AND THE SEX AND EMOTIONAL STATE OF THE SPEAKER. VOICEPRINTING USES SPEECH SPECTROGRAMS TO RECORD THIS INFORMATION. HOWEVER, CHANGING PHYSIOLOGY IN PERSONS, ALTERING STATES OF HEALTH, ATTEMPTS AT VOICE DISGUISE, AND SO ON, HAVE INTERFERED WITH THE ABILITY OF VOICEPRINTS TO ABSOLUTELY IDENTIFY SPEAKERS. EVEN WHEN A SPECTROGRAM MATCHING TECHNIQUE IS COMBINED WITH AURAL-PERCEPTUAL JUDGEMENTS BY THE OPERATOR, RESULTS ARE NOT ADEQUATE TO ESTABLISH THE VALIDITY OF VOICEPRINTING. A LABORATORY EXPERIMENT INVESTIGATING THE SENSITIVITY OF THE VOICEPRINT METHOD TO EFFECTS OF SPEAKER DISGUISE IS REPORTED. SUBJECTS READ A STANDARD PASSAGE TWICE, FIRST UNDER NORMAL SPEAKING CONDITIONS AND SECONDLY, WHILE DISGUISING THEIR VOICES. THE VOICEPRINT METHOD WAS APPLIED WITH THE FOLLOWING EXCEPTIONS: (1) HIGHLY SOPHISTICATED AUDITORS WERE USED, (2) THE AUDITORS DID NOT HEAR THE TALKER'S VOICE, AND (3) ALL JUDGMENTS WERE MADE IN A SEMI-CLOSED SET BASIS. THE METHOD FAILED TO PROVIDE CORRECT IDENTIFICATION AS MUCH AS 25 PERCENT OF THE TIME. THE PAPER SUGGESTS CRITERIA THAT A TECHNIQUE MUST MEET IF IT IS TO PROVIDE LAW ENFORCEMENT AGENCIES WITH A RELIABLE INVESTIGATIVE TOOL. REFERENCES ARE PROVIDED. SEE ALSO NCJ 38864, 52569, 57707-57709, AND 57711-57726. (AUTHOR ABSTRACT MODIFIED--PRG)