Hybrid Approach of Emotion Classifier

Abstract

Debate on the how emotion works?, is going on from last few decade. Emotions are brain states accompanied by corresponding bodily responses. These are the fundamental facts of the emotion. The conscious feeling we become aware of add to the emotional state. Emotions are, first and foremost, internal feelings we experience. and hence emotional expressions important for displaying our internal feelings. Human speech conveys linguistic messages as well as emotional information. Depending on acoustic parameters it is possible to measure multiple emotions. Discrete emotion theory and Dimensional theories have been introduced for emotions like Happy, sad, fear and, positive, negative respectively. In this paper we propose a hybrid model detecting type and class of emotion. SVM classifier approach is proposed for better results. Outcome of the model will compare with both Discrete and Dimensional theory base existing applications. With designing of hybrid model we hope that it will achieve better results. We also contribute a new dataset for emotion with Marathi and Hindi (Indian)database of speech.