CodeForge

▍HMM training algorithm

Hidden Markov models (HM m) traditional training methods of--Baum-W Elch algorithm can only be a local optimum model, thus affecting the Final recognition rate. CHM m, subparagraph k average method to obtain the original value can solve this problem, but the DHM m improvements are unlikely. An imp...

▍Speaker recognition based on LPCC and MFCC

MFCC parameters and speaker recognition LPCC parameters are the two most commonly used features of the parameters studied algorithm principle and LPCC MFCC parameter extraction and poorPoints cepstrum parameter extraction method, using MFCC, LPCC and the first order, second order difference as the c...