Facial Expression Recognition Based on the Difference of Statistical Features

In the methods of facial expression recognition based on statistical feature extraction, statistical features of facial expression images are correlative to shooting condition, illumination and the difference of images between subjects themselves (such as complexion), which will disturb the facial expression recognition. This paper points out a method of facial expression recognition based on the difference of statistical features (DSF). The differences of statistical features are obtained by subtracting statistical features of neutral expression from the statistical features of other expressions. By this way, we can eliminate partly these disturbances. Our experiments are performed on the databases of CED-WYU (1.0), COHN-KANADE and JAFFE. Our experiment results indicate that the recognition ratios of facial expression are heightened by this method