An Enhanced LBP Feature Based on Facial Expression Recognition

Because of excellent capability of description of local texture, Local Binary Patterns (LBP) have been applied in many areas. In this paper, we enhance the classical LBP method from three aspects for facial expression recognition: image data, extracting features and the way of combining all these features. At first, we adopt wavelet to decomposed images into four kinds of frequency images from which the features are extracted to increase original data. Then we extract LBP features with a new local and holistic way to make features more robust. At last, in order to use the extracted features more logical, we combine all data in an adaptive weight mechanism. All experiments are also proved that the proposed improvements in this paper have promoted the performance of facial expression recognition greatly.