Facial Expression Recognition Based On The Optimal Matching Of Feature-Classifier And DBN Decision

As the most intuitive way of expression of human emotions, facial expression and related research work have had a profound impact on health, business and family life etc. The thesis two aspects of facial expression recognition, one is feature extraction, the other one is expression classification. Improve the traditional Principal Component Analysis (PCA) algorithm, and put forward the extraction method to low space dimension. According to the complementarities of different feature and the difference of classification results, this paper proposes a fusion method of Multi-feature and Multi-classifier, Finally, basing on the Dynamic Bayesian networks, the model makes adaptive decisions for images that are still different to get the final recognition result. The main work is as follow:(1) Principal Component Analysis (PCA) can effectively extract global features from images and has advantages of dimension reduction. During the dimension reduction process, because of the comparatively concentration of eigenvalues, the dimension is still larger than the best. To solve this problem, this paper presents the optimal-sample PCA (OS-PCA) for dimension reduction. Through choosing the training samples and optimizing the covariance matrix, OS-PCA achieves the purpose of further dimension reduction. Because Discrete Cosine Transform (DCT) has robustness of light, as well as Local Binary Pattern (LBP) is effective in describing local texture features, this paper combines DCT and LBP features to make up for the limitations of OS-PCA in facial expression representation. In order to utilize the advantages of collaboration features and classifiers, this paper constructs a facial expression recognition model, which is based on three layers of the optimal integration of multiple classifiers. Firstly, the OS-PCA, DCT and LBP features are delivered into the model. Then, the model completes the optimal integration of multiple features and multiple classifiers. Via voting mechanism, the model makes adaptive decisions for images that are still different to get the final recognition result.(2) In view of the problem of high time complexity for decision fusion, this paper proposes a dynamic Bayesian classifier fusion model of multiple features and multiple classifiers for facial expression recognition. Firstly, basing on the training database, the model completes the optimal matching of Multi-feature and Multi-classifier and confusion matrix initializations. Then, the OS-PCA, DCT and LBP feature of test samples are put into the model, and the model can get three coarse classification results. Finally, basing on the experience of information and the established dynamic Bayesian network, the model makes adaptive decisions for images that are still different to get the final recognition result and updates the confusion matrix. The experimental results show that, in recognition rate and time complexity aspects, the model for facial expression recognition has higher reliability.