综合分析了语音和音乐的区别性特征 ,包括音调、亮度、谐度等感觉特征与 MFCC(Mel- Frequency CepstralCoefficients)系数等 ,提出一种 left- right DHMM(Discrete Hidden Markov Model)的分类器 ,以极大似然作为判别规则 ,用于语音、音乐以及它们的混合声音的分类 ,并且考察了上述特征集合在该分类器中的分类性能 .实验结果表明 ,文中提出的音频特征有效、合理 ,分类性能较好 This paper analyzes the distinguishing features of speech and music synthetically, including the sensory characteristics of pitch, brightness, harmonic and MFCC coefficients, and proposes a classifier of left-right DHMM (Discrete Hidden Markov Model) The maximum likelihood is used as the discriminant rule for the classification of speech, music and their mixed sounds, and the classification performance of the above feature set in this classifier is investigated. The experimental results show that the audio features proposed in this paper are effective, reasonable, Better performance