Features Extraction and Analysis of Disguised Speech Formant Based on SoundTouch

摘要

The identification of disguised voice is a difficult problem in forensic identification. Because the parameters of the formant have changed, there will be a great deviation from the original voice formant data, and the accuracy of speaker identification will be reduced or even unable to identify. In order to avoid investigation, criminals in cases often disguise their real voice, which brings great difficulties to case detection and voice identification. In the process of speaker recognition, the extraction of formant features is the key. In this paper, the camouflage voice generated by SoundTouch audio processing software is studied. After the formants of disguised voice are preprocessed, such as formant breakage cleaning, merge optimization, sequence adjustment and so on, the camouflage characteristics of electronic camouflage speech formants and the influence factors of different camouflage degrees on the camouflage characteristics of formants are extracted by regression equation. The influence of various camouflage methods on the parameters of formant is discussed.