|
Facial expression is a basic form of human expression of emotions, it plays an important role in the people’s exchange, not only can express human thoughts and feelings accurately, but also through facial expressions to understand each others’attitudes and inner world. Automatic facial recognition is premise of understanding the human emotions for computers, and it has unlimited potential in real life, such as human-computer interaction, affective computing, psychology, clinical medicine. Because of this special role, people do much more work in this area. The past 20 years, facial expression recognition method has been made very significant progress. However, due to facail expression recognition involves image processing, psychology, computer vision, artificial intelligence and other subjects, Because of this complexity and particularity, there are a lot of problems need to solve.This paper focuses on the accuracy of facial recognition. to carry out the exploration of key issues.Therefore, this paper’s main contents include the following aspects:First of all, on the Adaboost algorithm, the main study is how to solve multiclassification problem, weak classifiers would be constructed by Adaboost algorithm to generate a strong classifier. To solve the multi-class classification problem, we designed classifier by one-to-one mode, so the number of strong classifiers of Adaboost was k(k-1)/2 (k,number of categories).Secondly, on analysis and comparison of a variety of classificaion, the main work is on the nearest neighbor method, and the decision tree, by analyzing the principle of the two algorithms, and comparing them based on the characteristics of the different classifiers. For facial expression recognition problems, this two classifiers are used as the weak classifers in the Adaboost algorithm, and compares their performance.Then study the expression of feature extraction algorithms, the principle of Gabor filters is conducted in-depth analysis and research, analyzing and comparing the parameters of Gabor filter, in order to gain the most suitable parameters of the Gabor filters.Finally, we combined the Adaboost algorithm, decision tree classifier, Gabor features to study facial expression recognition, and used the JAFFE expression database and the Yale face image database to test. In the last, not only summarizes this paper’s work, but also clearly defined the direction and goals of future work.The proposed facial expression recognition based on Gabor feature and Adaboost can solve multi-class classificaion problems effectively. The results also show that this algorithm can improve the accuracy of identification and obtain better recognition results.
|