Facial expression recognition based on Gabor Wavelet transform and Histogram of Oriented Gradients

In order to get more effective expression features, this paper proposes an approach based on Gabor feature and Histogram of Oriented Gradients (HOG). Gabor Wavelet filter is first used as preprocessing stage for feature extraction. Handing the characteristics with a large number of dimensions, binary encoding (BC) is applied for dimensionality reduction. Dimensionality of the feature vector is reduced by using HOG algorithm. Experiments were performed on Cohn-Kanade facial expression database and the support vector machine classifier is used for expression classification. We obtained experimental results with an average recognition rate of 92.5%, which reveals that the proposed method is superior to other Gabor Wavelet transform based approaches under the same experimental environment.