Expression feature extraction based on difference of Local Binary Pattern histogram sequences

In this work, we present a novel facial feature extraction method for automatic facial expression recognition, which is based on difference of local binary pattern histogram sequences. First, the Local Binary Pattern (LBP) technique is used to extract facial texture features. Then the LBP Map is divided into non-overlapping rectangle regions with specific size, and histogram is computed for each region, which is concatenate to form histogram sequence called LBPHS. At last, the difference between the LBPHSes (DLBPHS) of the expression image and the neutral expression image is computed to form the final facial expression feature. The proposed feature extraction algorithm is tested on the JAFFE database and experimental results show that it is promising.