关键词: 完全局部二值模式（CLBP）, 有判别力的完全局部二值模式（disCLBP）, 表情识别, 最近邻

Abstract: To overcome high dimension and characteristic redundancy of CLBP features which result in the decrease of recognition speed and low recognition rate, this paper proposes an algorithm for facial expression recognition based on discriminative Completed Local Binary Pattern（disCLBP）. First of all, the facial expression sub-regions are isolated by a preprocessing step; then sub-regions and the global facial expression images are extracted by disCLBP, and different features are filtered for different expressions. After that cascade histograms are generated by associating the histograms of the filtered features of the images and the sub-regions. In the end, filtered features are classified by nearest neighbor classifier. Theaccuracy of the experiment on the facial expression database of CK is 97%.

Key words: Completed Local Binary Pattern（CLBP）, discriminative Completed Local Binary Pattern（disCLBP）, facial expression, nearest neighbor