Several Algorithms Based On Nonlinear Dimensionality Reduction Of Facial Expression Recognition Research

Facial expression is an important body language. Studies indicate that, in our daily life, only 7 percent of emotion information is expressed by oral language, but 55 percent of emotion information is expressed by facial expression. The major work of facial expression recognition is using computer to extract facial features of all expression images. Then according to the difference of features, each image is classified to one of seven the classes different expression. This makes computer understand the expression states from the classify result and realized Human-Computer Interaction. Despite many progresses have been got in FER, but in existing environment, the interference of illumination, posture, noise, masking, and so on, which make facial expression recognition technology still needs some more researches before it realizing practical applications.In this paper, we analyzed the progress of domestic and international facial expression recognition technology in recent years and discussed several issues on facial expression recognition by machine. The application of nonlinear dimensional reduction method in facial expression recognition is detailed introduced and a series of experiments are carried out. The work in this paper mainly includes the following several respects:1. We have introduced the research background of facial expression recognition and reviewed the FER techniques for feature extraction and classification.2. We have introduced some common dimensionality reduction techniques. Comparing with each other, we find their merits and drawbacks of each algorithm.3. A facial expression recognition method based on SNE and SVM was studied. It explicitly introduced the algorithm of SNE and SVMs. The proposed algorithm use SNE as a dimensionality reduction toolkit and SVM as a classifier. By applying this algorithm to FER with JAFFE database, it recorded the highest performance of 65.7 percent FER rates, which higher than some traditional FER algorithm, such as PCA+SVM and LDA+SVM.4. A facial expression recognition method based on GPLVM plus SVM was studied. It explicitly introduced the theory of GPLVM. The algorithm use GPLVM for reducing the high dimensional data of facial expression images into a relatively low dimension data and apply support vector machine (SVM) as classifier for the expression classification lately. By applying this algorithm to Japanese Female Facial Expression (JAFFE) database for facial expression recognition, we find that the proposed new algorithm has a better performance than the traditional algorithms, such as PCA and LDA etc. The maximum recognition rates of facial expression recognition in non-given person of GPLVM+SVM on JAFFE database are 65.24 percent.5. A facial expression recognition method based on SPE plus SVM was studied. This algorithm uses stochastic proximity embedding (SPE) for reducing the high dimensional data of facial expression images into a relatively low dimension data and then uses support vector machine (SVM) as the classifier for the expression classification afterwards. Applying this algorithm to Japanese Female Facial Expression (JAFFE) database for facial expression recognition, we find that the proposed new algorithm has a better performance than the traditional algorithms, such as PCA and LDA etc., The maximum recognition rates of facial expression recognition in non-given person of SPE+SVM on JAFFE database are 69 percent. Besides, due to the features extracted by SPE, the algorithm of SPE+SVM attains higher FER performance in a relative low dimensionality of the input data, and behaviors better than PCA+SVM in recognizing some ambiguous expressions, which can interpret as SPE algorithm has a stronger robustness than PCA and enables it work well on fuzzy environments.