METHOD FOR PERSON IDENTIFICATION USING ARBITRARY SPEECH RECORDS

摘要

FIELD: crime investigation. SUBSTANCE: method involves converting signals of unknown person and person under investigation into digital form, storing digital images in computer memory, subjecting this images to acoustic and linguistic analysis, calculation of acoustic characteristics of these persons and comparison of vectors of characteristics using results of preliminary analysis (signal-noise ratio of constituents of amplitude-frequency spectrum and frequency band). This results in possibility of combination of acoustic and linguistic analysis, use of statistic criteria for decision making, flexible tuning to quality and duration of records under investigation, possibility to tune decision rules for specific persons, use of computer database of sound standards in linguistic analysis. EFFECT: increased efficiency of identification in presence of noise and distortion. 1 dwgn