The influence of fundamental frequency on speaker recognition system

The impact of fundamental frequency of MFCC usually is neglected in the application and research of speech signal based on feature MFCC. Analysis shows that the fundamental frequency affects the accurate description of channel characteristics, thus affect the performance of speaker recognition system, so a new speech signal feature named SMFCC (MFCC smoothing) is proposed based on the smoothed amplitude spectrum. Experiments on the YOHO speaker recognition database show that the performance of SMFCC is much more precise than MFCC. Especially in the female speaker data set, the performance of SMFCC is obvious, and SMFCC has better robustness.