The Research of Forensic Voiceprint Identification Based on WMFCC
Guorui Feng, Xu Chang · 2019
The characteristic comparison of formants is an important method for voice identification. In order to solve the key problems of voice identification, a weighted Mel Frequency Cepstrum Coefficient (WMFCC) which can be used for formats extraction from the voice signal is proposed in this paper. The WMFCC coefficient which contains the main components of the frequency spectrum can be obtained by analyzing short-time voice signal. The candidate formants can be obtained from the formant position of spectral envelope parameters. The accurate formant location would be fixed by analyzing the main components of signal spectrum. The resonance formats analysis of Nasal vowels; the vowel resonance formants contrast and Long Term Mean Power Spectrum Analysis were used for the Voiceprint identification. The experiment shows fewer errors of the formants based on WMFCC were smaller, and it shows better robustness in noisy environment.