On the person-independent facial expression recognition, the utilization rate of the facial expression texture is not high. Facing with the problem of the person-independent face, this paper proposes a method about facial expression recognition based on the improved weighted local binary pattern (LBP) and sparse representation. In order to use the local texture information of the facial organs effectively, first it uses the improved weighted LBP operator to extracting the local texture feature, the extracted features to construct the training samples, and classified via the sparse representation last. Experimental results show a better performance on the JAFFE and CK database.