Abstract—In this study, an automated human emotion recognition system based on contourlet transform is presented. The various human emotional stats such as surprise, sadness, disgust, anger, fear, happiness and neutral are analyzed. The emotional states are identified by the motion of the muscles in the skin of the human face. In order to classify the emotional state, the patterns generated by the muscles motions should be represented efficiently. In the proposed approach, these patterns are characterized by contourlet transform. The high dimensional contourlet transformed face image is translated into low dimensional space in the feature extraction stage. This can be achieved by taking the average of contourlet coefficients in each sub-band and fed into the classifier. The number of features used for classification depends on the level of decomposition used while transforming the face image by contourlet transform. The emotional states are classified by nearest neighbor classifier and the classifier performance is analyzed by various distance measures; euclidean, city block, cosine and correlation efficiently. The results show that the proposed approach based on contourlet transform produces over 80% recognition accuracy on JApanese Female Facial Expression (JAFFE) database.

Index Terms—contourlet transform, nearest neighbour classifier, facial expressions, emotion recogniton

Cite: R. Suresh, S. Audithan, and P. Elakkiya, "Contourlet Tranform Based Human Emotion Recognition System," International Journal of Signal Processing Systems, Vol. 2, No. 1, pp. 7-11, June 2014. doi: 10.12720/ijsps.2.1.7-11