Facial Expression Recognition Based on Adaptive Weighted Fusion Histograms

摘要

In order to improve the performance of expression recognition, this paper proposes a facial expression recognition method based on adaptive weighted fusion histograms. Firstly, the method obtains expression sub-regions by pretreatment, and calculates the contribution maps (CM) of each expression sub-region. Secondly, this method extracts Histograms of Oriented Gradient by the Kirsch operator and extracts histograms of intensity by centralized binary pattern (CBP), then the paper fuses the Histograms by parallel manner and the fused histograms are weighted by CMs. At last, the weighted fused histograms are used to classify by the Euclidean Distance and the nearest neighbor method. Experimental results which are obtained by applying the proposed algorithm and the Gabor wavelet, LBP, LBP+LPP, Local Gabor and AAM on JAFFE face expression dataset show that the proposed method achieves better performance for the face expression recognition.