Research And Implementation Of Multi-lingual Speech Emotion Recognition

Social perceptions and judgments of emotion are ubiquitous and are deemed important to human survival. With the development of science, the processing and analysis of emotion was already became an important research direction in the field of Artificial Intelligence. In traditional speech signal processing, the computer can only recognize the information about symbol, the hidden information related with emotion and mood is dropped without been noticed. The development of speech emotional analysis will improve and enhance the computer's emotional intelligence—the ability to recognize a user's affective states, to became more human-like, more effective and more efficient.In the beginning of this paper, the research background and history of affective speech processing are reviewed. We chose proper emotional acoustic parameters to this paper, after deeply researching. Time parameters, pitch parameters, energy parameters and formant parameters are used as prosodic parameters in the statistical analysis, which are extracted from a emotional speech database of German. And we carried out three methods to recognize emotion states, which are Principal Component Analysis(PCA), Fuzzy Mode Recognition and Gaussian Mixture Model(GMM). An improvement has made to Fuzzy Mode Recognition, we use the idea of maximum separability from PCA to allot different value to each prosodic parameter. And this method has made a little improvement to the recognition rate.