A Novel Approach for Bimodal Human Emotion Recognition

In this paper six basic (Fear, Angry, Happy, Neutral, Sad, Surprise) human emotions are recognized by bimodal fusion techniques. The emotion features are extracted both from facial images and speech signal. All the features collected from both the domain are fused using Discrete Wavelet Transform and Logistic Regression Model to generate single feature set. Both the fused feature set classified by Back Propagation Network. It is observed that the accuracy ofDiscrete Wavelet Transform is 91.67% which is better than Logistic Regression of accuracy 89.17%.