Abstract: Non-stationary characteristics of speech signal under the different emotions are especially obvious. Traditional MFCC can only reflect speech static features, while EMD can describe non-stationary characteristics of speech signal precisely. In order to extract the non-stationary features of emotional speech, the improved MFCC steps are proposed including EMD decomposition into IMFs, Mel filtering, logarithm and DCT. The improved MFCC is adopted as the new feature with SVM to recognize four speech emotions consisting of happy, angry, bored and fear. Simulation results demonstrate that the recognition rate of the improved MFCC is 77.17%, and in different SNRs, the recognition rate can be increased by 3.26%.

Key words: speech emotion recognition, Mel-Frequency Cepstral Coefficients（MFCC）, Empirical Mode Decomposition（EMD）, Support Vector Machine（SVM）

摘要： 人在不同情感下的语音信号其非平稳性尤为明显，传统的MFCC只能反映语音信号的静态特征，经验模态分解能够精细地刻画语音信号的非平稳特性。为提取情感语音的非平稳特征，用经验模态分解将情感语音信号分解为一系列固有模态函数分量，通过Mel滤波器后取其对数能量，进行DCT反变换后得到改进的MFCC作为情感识别的新特征，采用支持向量机对高兴、生气、厌烦和恐惧等四种语音情感识别。仿真实验结果表明：改进的MFCC识别率达到77.17%，在不同的信噪比下，识别率最大可提高3.26%。

关键词: 语音情感识别, Mel频率倒谱系数, 经验模态分解, 支持向量机