Spontaneous emotion recognition for Marathi Spoken Words

Abstract

In this paper analysis of emotion recognition from Marathi speech signals by exploring several patterns for feature extraction techniques and classifiers to classify speech utterance according to their emotion contains. In this paper several method are extracting feature from speech signal to estimation of energy, intensity and pitch contour using Mel Frequency Cepstral Coefficient (MFCC). These feature parameters are extracted from Marathi speech Signals depend on speaker, spoken word as well as emotion. Gaussian mixture Models (GMM) is used to develop Emotion classification model. Each subject/Speaker has spoken 7 Marathi words with 6 different emotions that is 7 Marathi words are Aathawan, Aayusha, Chamakdar, Iishara, Manav, Namaskar, Uupay and 6 emotions are Angry, Happy, Sad, Fear, Neutral/Normal, and Surprise. This system is used for emotion recognition in Marathi Spoken Words by applied feature extraction techniques as MFCC and classification techniques as GMM. We got 83.33 % average accuracy rate and 16.67% average confusion rate of our system. For Male we got average accuracy rate is 85% and for female 81.66 %. This is the overall accuracy rate of our Emotion Recognition for Marathi Spoken Words (ERFMSW) system.