The Recognition Of Facial Expression Based On SIFT Algorithm

In recent years, the field of naturalized and intelligent human-computer interactionresearch and the emotional intelligence field have achived a lot of remarkableachievements. In these areas, a key technology of how to get people’s inner feelings hasbeen paid a lot attention. Expression is an important form to display people’s innerfeelings, which can deeply reflect changes of the people’s inner state, therefore, thefacial expression recognition technology attracts wide attention from researchers.This paper firstly discribes related theory and technology of the facial expressionrecognition in recent years, and puts theory into practice.The self adaptive thresholdingmethod and Adaboost method based on Haar features are used to detect faces. Secondly,Scale-invariant feature transform (SIFT) method is used to extract features fromexpression images, meanwhile, in order to improve the accuracy of recognition timely,this paper extracts features from the training samples, in which15representative featurepoints will be selected to represent an expression image according to Fisher criterion.Thirdly, discriminative scale-invariant feature transform (D-SIFT) and the weightedmajority voting (WMV) classifer are employed to get the final result.Finally, this paper implements a facial expression recognition system based on themethods that have been refered above. It takes three steps to gain this purpose, they aredetecting face, extracting features and recognizing facial expression. At the same time,the accuracy and effectiveness of the methods could be verified by experiments basedon this system.