Abstract：A facial expression recognition algorithm based on the feature extraction method of LBP was introduced. Firstly, face images were divided into local patches, and the information of facial action units was extracted to compose the compositional features of facial expressions. The coarse position of eyes and mouth was located. Secondly, the feature extraction method of local binary patterns ( LBP) was adopted to decompose the images into sub-blocks, and each sub-blocks histogram was attained. Then the LBP histogram of overall features and LBP histogram of local features were combined into the LBP histogram of a whole image. Finally, the LBP histogram of a whole image was sent to an embedded hidden Markov model ( EHMM) as an initial vector, and the observation sequence was formatted. Through testing seven basic expressions in JAFFE face database, the experiments show that the algorithm can enhance facial expression recognition effectively.

Keyword：facial expression recognition;feature extraction;feature combination;LBP histogram;