Research of speech emotion recognition based on emotion features classification
基于情感特征分类的语音情感识别研究

Because the speech signals were highly real-time uncertainty, this paper proposed evidences' trust entropy and dynamic prior weights to improve the basic probability function of traditional D-S theory. As the emotion recognition result was not the same by emotion features in different emotions, it presented a classification method of emotion features. In order to realize the fine-grain speech emotion recognition, it used the recognition data of different classification and the improved D-S theory to realize the emotion recognition based on multi-classification emotion features. The improved D-S theory is proved to be effective by simulation. And comparing simulation results show that the multi-classification emotion features are effective and stability.