ROLE OF IDIOSYNCRACIES IN LINGUISTIC STRESSING CUES, AND ACCURATE FORMANT ANALYSIS, IN SPEAKER IDENTIFICATION (FROM CARNAHAN CONFERENCE ON CRIMINAL COUNTERMEASURES PROCEEDINGS, 1976, BY JOHN S JACKSON - SEE NCJ-57707)

• NCJ Number: 57709
• Author(s): A D C HOLDEN; J Y CHEUNG; Y K GULUT
• Date Published: 1976
• Length: 7 pages
• Corporate Author: University of Kentucky
• Publication Format: Document
• Language: English
• Country: United States of America

Annotation

THIS PAPER INTRODUCES NEW FEATURES TO INCREASE THE RELIABILITY OF EXISTING SUCCESSFUL SPEAKER IDENTIFICATION SYSTEMS.

Abstract

SUCCESSFUL SPEAKER IDENTIFICATION DEPENDS UPON FEATURES, DERIVED FROM THE SPEECH SIGNAL, WHICH REFLECT THE IDIOSYNCRASIES OF THE INDIVIDUAL SPEAKER. A NEW PITCH-SYNCHRONOUS ANALYSIS TECHNIQUE IS PRESENTED WHICH PROVIDES HIGHER ACCURACY WITH LESS VARIANCE IN FORMANT FREQUENCY, BANDWIDTH, AND AMPLITUDE FORMATION; CORRECTS THE SPECTRAL BALANCE; FOLLOWS FAST FORMANT TRANSITIONS; PERFORMS WELL WITH NASAL AND HIGH-PITCHED SOUNDS; AND CANCELS PERIODIC NOISE. EXPERIMENTAL EVIDENCE SHOWS THAT INDIVIDUAL SPEAKERS USE FIVE ACOUSTIC CUES FOR LINGUISTIC STRESS, WITH VARYING DEGREES OF EMPHASIS ON EACH CUE, DEPENDING ON THE INDIVIDUAL. A MODEL ALSO IS GIVEN WHICH GENERATES THE STRESS PATTERNS IN A SENTENCE IN A MANNER CLOSELY FOLLOWING THAT USED BY HUMAN SPEAKERS. THE MODEL EMPLOYS PARAMETERS RELATED TO THE 'ANTICIPATORY' AND 'RECENCY' INFLUENCES OF 'SYLLABLES WITH MAJOR STRESS' ON 'SYLLABLES WITH MINOR STRESS.' THE PARAMETER VALUES ARE CHARACTERISTIC OF THE INDIVIDUAL SPEAKER. FIGURES, TABLES, AND REFERENCE NOTES ARE INCLUDED. SEE ALSO NCJ 38864, 52569, 57707-08, AND 57710-57726. (AUTHOR ABSTRACT MODIFIED--PRG)