Facial Expression Recognition Based on Gabor Wavelet Transform and Relevance Vector Machine

摘要

Facial expression recognition plays an important role in intelligent human-machine interaction. This paper proposes an effective algorithm for recognition of six basic facial expressions. The algorithm utilizes Gabor wavelet transform to get expression features, and adopts local nonuniform feature point extraction, Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT) to solve the problem of high dimension. The facial expressions are classified by Relevance Vector Machine (RVM), which its classification performance is better than the Support Vector Machine (SVM). At last, according to the recognition results of the six basic expressions, an optimum decision scheme, which based on the tow-against-two classification method, is designed to realize a quick and accurate expression classification. The comparative experiments between RVM and SVM demonstrate that the proposed algorithm can further improve the recognition accuracy and achieve better generalization performance.