Facial Expression Recognition Method Based On Selective Ensemble

With the rapid development of artificial intelligence, human feel so untouched that computer can have intelligence like human and have feelings like we do. For traditional method of facial expression recognition, researchers often use single classifier or integrate multiple classifiers to recognize emotions. The traditional facial expression recognition method has low recognition rate, large storage expenses that lead to high forecast cost. As selective ensemble method can solve these problems, it has been gradually applied to facial expression recognition. It brings a new opportunity and challenge for the facial expression recongnition technology by combining selective ensemble and expression recongnition method.The thesis studies about selective ensemble theory based on the traditional methods of facial expression recognition, and mainly focus on selective ensemble, semi definite programming theory, simple particle swarm optimization algorithm based on extremum disturbed and parallel feature fusion.For making up for the defects of traditional facial expression recognition, a new facial expression recognition based on selective ensemble is proposed. The selection of based classifications would convert into a semi-definite programming problem in the new facial expression recognition. And the simple particle swarm optimization algorithm based on extremum disturbed would be used to solve the semi-definite programming problem. The experimental results show that, in comparison with the traditional facial expression recognition methods and other selective ensemble for expression recognition methods, the proposed method can improve the recognition rate.Furthermore, based on above proposed, another new facial expression recognition method based on parallel feature fusion is proposed. This method extracts two kinds of features, uses principal component analysis method and parallel features fusion method to reduce the features' dimensionality reduction and construct the combination features.Finally, a facial expression recognition system based on selective ensemble is designed and developed in this thesis. This system recognize the facial expression with following steps: feature extraction and feature reduction, training base classifier, selection and the integration of multi classifier. The actual test indicating that the system is able to effectively identify the face of offline face.