With the development of computer technology, it is a research topic currently attracting much attention that how to identify the emotional state of the speaker automatically from speech. As a single classifier in the limitation of speech emotion recognition, we designed three kinds of classifier based on Hidden Markov Models (HMM) and Artificial Neural Network (ANN) for the four emotion of angry, sadness, surprise, disgust in this paper . Then DS evidence theory was proposed to execute decision fusion among the three kinds of emotion classifiers for a good emotion recognition result. Based on the Berlin database of emotional speech, DS evidence theory was confirmed a feasible method to significantly improve the accuracy of the speech emotion recognition, and the average recognition rate of fore emotion states has reached 83.86%.
展开▼