Facial Expressions Recognition Based on ULBP and Edge Map

This paper presents an algorithm for automatic facial expressions recognition. The algorithm is based on the Local Binary Patterns (LBP). The proposed algorithm first calculates the horizontal and vertical edges of the input image that gives important information of facial expressions. Facial expressions are generated by movement of muscles that are better represented in the form of edges. Next LBP histograms are extracted from both the edge maps after dividing the edge image into blocks. In addition, in order to reduce the dimensions of LBP histogram only uniform patterns (ULBP) are used, which reduces the computational complexity of facial expressions recognition. The proposed algorithm is tested on standard facial expressions databases TFEID & JAFFE. The experimental result shows that the proposed algorithm can obtain average recognition rate of 80% on TFEID images when using fixed and adaptive weights. In case of JAFFE images, recognition rate is 80% and 74% respectively.