International Journal of Computer Science Issues

More than a traditional journal...

Emotion recognition helps to recognize the internal expressions of the individuals from the speech database. In this paper, Dynamic time warping (DTW) technique is utilized to recognize speaker independent Emotion recognition based on 39 MFCC features. A large audio of around 960 samples of isolated words of five different emotions are collected and recorded at 20 to 300 KHz sampling frequency. Training and test templates are generated using 39 MFCC features. In the proposed work, we have extracted the MFCC coefficients from the speech database and DTW is used to store a prototypical version of each word in the vocabulary and compute incoming emotion with each word. For the classification of emotions SVM is used. The experimental results are provided using MFCC, Delta Coefficients (#8710;MFCC) and Delta Delta Coefficients(#8710;#8710;MFCC) . It is proposed that higher recognition rates can be achieved using MFCC features with DTW which is useful for different time varying speech utterances.