Research on Tibetan Lhasa Dialect phonetic feature extraction technology based on LDA-MFCC

摘要

Tibetan feature extraction algorithm is the most critical link in Tibetan speech recognition system. According to Tibetan Lhasa Dialect phonetic and pronunciation features, the Mel frequency cepstrum coefficient (MFCC) feature extraction algorithm is established in this paper based on the simulation of human auditory system, and extracted feature data are compressed through the LDA information compression algorithm. Recognition rate and operational efficiency are improved while dimensions are reduced. A feature extraction algorithm of LDA-MFCC based on Tibetan Lhasa dialect phonetic features is finally concluded.