Abstract:

Aiming at the deficiency of the traditional expression recognition methods for complex situation, we propose a novel person-independent facial expression recognition method. We first calculate the histogram of orientation gradient for each facial expression image and extract Haar wavelet features from them. Then by connecting the two different characteristics obtained above, we get the whole image features. By using the multi-class SVM classifier in each layer we finally achieve facial expression classification and recognition. Simulations of facial expression recognition on the Japanese Female Facial Expression (JAFFE) database show that the classification accuracy reaches up to 87.9%, and the average time consumption rises to 10.2966s. The results of comparison experiments show that the new algorithm has higher recognition accuracy rate, less time consumption and stronger robustness.

Key words: histogram of oriented gradient, Haar wavelet, decision tree, SVM