Speech Emotion Recognition Based on Integration of Multiple Classifiers

A new method of speech emotion recognition via integration of multiple classifiers is proposed for improving speech emotion classification rate. Based on extracting prosody, voice quality and MFCC feature parameters from emotional speech, three kinds of classifiers including Bayes.net, K-Nearest Neighborhood (KNN) and Radial Basis Function (RBF) neural network are utilized to construct an ensemble classifier so as to realize recognizing the seven main speech emotion in Berlin emotional speech database like anger, joy, sadness and neutral, fear, bore and disgust. Computer simulation results show that the ensemble classifier can achieve average correct rate of 71.4019%for speech emotion classification, which is superior to the single classifier.