NCJ Number
56188
Date Published
1977
Length
5 pages
Annotation
BECAUSE THE EFFECTS OF DISGUISED SPEECH REMAIN AN UNKNOWN FACTOR IN VOICE IDENTIFICATION EFFORTS, AN ACOUSTICAL ANALYSIS OF 23 ADULT MALES WAS UNDERTAKEN TO COMPARE THE PARAMETERS OF NORMAL AND DISGUISED SPEECH.
Abstract
ABILITY TO DISGUISE THE VOICE HAS CREATED A MAJOR OBSTACLE TO THE DEVELOPMENT OF SATISFACTORY VOICE IDENTIFICATION SYSTEMS FOR USE AS A CRIME COUNTERMEASURE. ONE REASON FOR THIS OBSTACLE IS THE SPARSITY OF RESEARCH DATA ABOUT VOICE DISGUISE. IN COMPARING THE ACOUSTIC DATA FROM DISGUISED SPEECH WITH THAT OF NATURAL SPEECH IN ADULT MALES, THE 23 SUBJECTS READ A PASSAGE NATURALLY AND AGAIN USING ANY FORM OF DISGUISE THEY CHOSE. EACH READING WAS DIVIDED INTO ITS SPECTRAL/TEMPORAL COMPONENTS. SPECTRUM VARIATIONS OCCURRED IN SPEAKING FUNDAMENTAL FREQUENCY, CENTER FORMANT FREQUENCIES, AND IN FORMANT BAND WIDTHS. DURATION WAS GENERALLY GREATER FOR DISGUISED VOICES WITH PROLONGATION OF CONSONANTS CONTRIBUTING MOST TO THIS DIFFERENCE. THESE RESULTS INDICATE THAT ACOUSTICAL COMPONENTS OF SPEECH CAN BE SIGNIFICANTLY ALTERED BY PERSONS SPEAKING WITH A DISGUISED VOICE. DEVELOPMENT OF EFFECTIVE VOICE IDENTIFICATION TECHNIQUES MUST, THEREFORE, INCLUDE CORRECTION FACTORS FOR THESE ALTERATIONS OR UTILIZE ONLY COMMONALITIES THAT MAY BE PRESENT WITHIN THE SPEECH SIGNAL. EXISTING METHODS HAVE HAD LITTLE SUCCESS IN ACHIEVING THESE REQUIREMENTS. TABULAR DATA AND REFERENCES ARE PROVIDED. (AUTHOR ABSTRACT MODIFIED--KBL)