Abstract
The MFCC and LPCC is the speaker recognition feature parameters to be used, the speech of analysis often use these feature, therefore, we need a parameter extraction algorithm and the principle of MFCC and LPCC. The reason is that w e need to understand the technological process. We are modeling to the speaker, the traditional vector quantization model increase the differential equation. The formation of a series of consecutive words distribution vector quantization model, at the same time, extracting the characteristic parameters of Mel frequency cepstrum coefficient and differential speaker that consists of linear prediction cepstrum coefficient combination, we conduct a text about dependent speaker recognition. The algorithm show mixing MFCC and LPCC extraction, based on keeping the response time of the system did not significantly increase the rate the model has been improved to be a certain extent.