关键词: 表情识别, 特征融合, 局部二值模式, 稀疏表示

Abstract: In order to effectively represent facial expression feature, a novel method fusing Local Binary Pattern（LBP） and local sparse representation is proposed for facial expression representation and recognition. The face image is divided into non-uniform local regions based on face features, and the LBP features are calculated for every facial local region. For depicting facial local texture exactly and integrating facial local features, a facial local reconstruction method based on sparse representation is designed. According to the influence on the facial regions for expression, a weighted fusion algorithm is presented to collect all the local reconstruction residuals and address the expression recognition task. The experiments tested on JAFFE2 face expression database demonstrate the proposed method is feasible and robust to facial expression.

Key words: expression recognition, feature fusion, Local Binary Pattern（LBP）, sparse representation