Emotional Feature Analysis and Recognition from Vietnamese Speech

In this paper we studied the emotion recognition from Vietnamese speech signal, and established a Vietnamese emotional speech database. Two male subjects and two female subjects whose native language is Vietnamese participated in the acting of emotional speech. Through a listening test by multiple listeners the emotional data is selected and a basic Vietnamese speech emotion database is achieved, which may serve as a data foundation for future cross-language study. Based on the collected emotion data we extract the basic acoustic features and construct the static emotional features which are used for modeling and recognition. Emotion model is built and tested using Gaussian mixture models, and experimental results show that the emotion recognition system proposed in this paper successfully detects several basic emotions from Vietnamese speech. In the future work a further study on the cross-language emotional feature analysis and recognition is still needed.