Face Expression Recognition Based on Feature Fusion

Abstract

The traditional Gabor transform need calculate convolution with whole image when extract expression feature. So, the feature dimensions are very large and computation complexity is very high. Though some improved methods can reduce the feature dimensions and computing cost by manually marking out the feature points, but those methods need intense human intervention and can not meet to the need of automatic recognition. In this paper, a new face expression recognition method are proposed based on feature fusion which combines the Gabor transform and ASM automatic feature orientation technology. Firstly, ASM technology is used to locate the feature point of a face shape. And then Gabor feature of feature point are extracted by using Gabor transform. Finally, the two feature sets are fused to implement the face expression recognition. The experimental results show that the proposed method can effectively utilize the local texture information and global shape information of expression and get better recognition effect.