Research On Facial Expression Recognition Method Based On Multi-feature Fusion

Facial expression recognition based on digital image processing and pattern recognition is an important technology in computer vision.When a single feature form is used to recognize the expression images with different illumination,direction or scale,it cannot effectively extract the expression features of various forms,and the recognition accuracy obtained is relatively low,so it cannot be applied to some applications with high requirements.In order to extract facial features more comprehensively,multi-feature fusion of facial expression is taken as the main research content in thesis to improve the robustness and applicability of facial expression recognition algorithm.The main research contents are as follows:The facial expression feature extraction method based on LBP algorithm is studied.The classic LBP operator,circular LBP operator and improved LBP operator are used to extract features from face images respectively.The facial expression recognition rate effect of classic LBP algorithm and improved LBP algorithm in JAFFE database and CK+ database is compared and analyzed.Experimental results show that the improved LBP algorithm has high recognition accuracy and short running time.This thesis studies the facial expression extraction method based on HOAG algorithm to solve the problem that the classical HOG algorithm does not consider the role of the central pixel.The effect of expression recognition rate of HOG algorithm and HOAG algorithm in JAFFE database and CK+ database is compared and analyzed.The experimental results show that HOAG algorithm has high recognition accuracy and short running time.Since a single feature form cannot effectively extract salient features of various forms in facial expression recognition of face images with illumination interference and large scale or angle gap,thesis proposes a facial expression recognition method based on multi-feature fusion according to the above research results.The improved LBP algorithm is used to extract texture features of face images,and the HOAG algorithm is used to extract local facial features of eyes,eyebrows and mouth,and the above features are fused together in series form in the feature layer for facial recognition.The method is tested on JAFFE database and CK+ database.The experimental results show that the recognition efficiency of CK+ database and JAFFE database can reach 98.6% and 94.17% respectively on SVM classifier.Compared with single feature and existing facial expression feature extraction methods,the method in thesis has a higher correct recognition rate.