Facial Expression Recognition Based On Biologically Inspired Features

Facial expression recognition is a cross-subject of artificial intelligence, com-puter vision, image processing, physiology and psychology, and has become a hottopic in pattern recognition and artificial intelligence. Although many algorithmsfor facial expression recognition has been presented up to now, facial expressionrecognition system is still in the stage of exploration, each method has its limi-tations. How to improve the accuracy and automatic degree of facial expressionrecognition is still an issue that is worth studying. Because Biologically InspiredFeature (BIF) has shown its superiority in human age estimation, gender and racerecognition, etc, the application of BIFs to facial expression recognition is focusedon in this paper. The main contributions are as follows:1. The research background of the facial expression recognition is introduced.A survey of feature extraction and classification for the purpose of facial expres-sion recognition is conducted.2. An algorithm for facial expression recognition based on BIF and SVM ispresented and implemented. First, BIFs are extracted from facial images, thenfeature dimension is reduced by PCA and LDA methods, finally SVM is used tofacial expression classification. Also, a similar algorithm based on Histogram ofOriented Gradient (HOG) is presented and implemented.3. Both algorithms are tested on Japanese Female Facial Expression Databaseand compared with other existed methods in current literatures. The experimentresults show that our methods have good performance and robustness.