关键词: 隐马尔可夫模型, 语音自动切分, Mel频率倒谱系数, 小波子带分解

Abstract: Two new feature parameters based on wavelet packet transform were proposed intend of MFCC. According to the difference of decorrelation method, the two feature parameters were named as subband based cepstral parameters (SBC) and wavelet packet parameters (WPP). The tests indicate that SBC and WPP achieve better performance than MFCC.

Key words: HMM, automatic speech segmentation, MFCC(Mel-Frequency Cestrum Coefficient), wavelet packet transform