Facial Expression Recognition under Partial Occlusion Based on Gabor Multi-orientation Features Fusion and Local Gabor Binary Pattern Histogram Sequence

In this paper, we propose a novel facial expression recognition method under partial occlusion based on Gabor multi-orientation features fusion and local Gabor binary pattern histogram sequence (LGBPHS). Firstly, the Gabor filter is adopted to extract multi-scale and multi-orientation features. Secondly, the Gabor magnitudes of different orientations in the same scale will be fused according to the fusion rule in this paper and then the fusion features are further encoded by using the LBP operator. Finally, the fused image is divided into several non-overlapping rectangle units with equal size, and the histogram of each unit is computed and combined as facial expression features. The proposed method is robust to partial occlusion and better recognition rates are achieved in JAFFE database with eyes occlusion and mouth occlusion. Experimental results show that the method is effective to facial expression recognition under partial occlusion.