Use of Multiple Classifier System for Gender Driven Speech Emotion Recognition

This paper proposes a system that allows recognizing a person's emotional state with the help of recording audio signals. This system is able to recognize four emotions (anger, happiness, sadness and neutral) This emotion recognition technique is mainly composed of two subsystems as - 1) gender recognition (GR) and 2) emotion recognition (ER). It has been proved experimentally that the performance of emotion recognition increases because of the apriori knowledge about gender of the speaker. Traditional approach shows that selection of proper and unique features of speech signals reduces the unnecessary calculation complexity. Recently use of combination of two or more different classifiers is emerging trend in the classification field. As HMM is the best training algorithm and SVM is the best classification algorithm, proposed technique makes use of hybrid of HMM and SVM classifiers to get best results.