Emotions recognition by speechand facial expressions analysis

In this paper, we propose a bimodal emotion recognition system using the combination of facial expressions and speech signals. The models obtained from a bimodal corpus with six acted emotions and ten subjects were trained and tested with different classifiers, such as Support Vector Machine, Naive Bayes and K-Nearest Neighbor. In order to fuse visual and acoustic information, two different approaches were implemented: feature level fusion and match score level fusion. Comparative studies reveal that the performance and the robustness of emotion recognition systems can be improved by the use of fusion-based techniques. Further, the fusion performed at the feature level showed better results than the one performed at the score level.