The Research Of Facial Expression Recognition Based On Mixed Feature Extraction And Decision Tree Algorithm

Facial expression recognition, as a new research direction in recent years, has gradually become a hot research topic in the field of human computer interaction. The difficulty of the facial expression recognition lies in the intersection of physiology, psychology, image processing, pattern recognition and machine vision. With the development of its research, we found that each link of the expression recognition, including image preprocessing, feature extraction, sample feature processing, can affect the final effect of facial expression recognition. And in the field of image processing, C4.5 decision tree algorithm has advantages including intuitive, low complexity of the algorithm. So it has the conditions of facial expression classification, for the above content, this paper launched the following research:1、For the single feature extraction can reflect the expression of the characteristics of a local or a single character, but not fully reflect the general characteristics of facial expression, in this paper, we use a hybrid feature extraction method based on local features and global features including basic geometric features, texture features based on the co-occurrence matrix and Hu invariant moments.2、In order to eliminate the influence of dimensional difference and the difference of numerical value,the sample data of training base should be standardized and visualized. So we use Fisher Linear discriminant to reduce the feature dimension we extract and use feature histogram normalization to normalize the data we reduced to further improve the recognition rate.3、In view of the fact that the C4.5 classification has been widely used in the field of image processing, the complexity of the algorithm is low, the classification accuracy is high, and the recognition rate is fast. Also, the researcher pretend to use the C4.5 classifier to predict and identify the image. So, We propose that the C4.5 classifier is applied to facial expression recognition and to improve the classifier based on L’Hospital Rule, the algorithm in classification complexity and speed can be further improved. Consider the problem that the decision tree is tend to overfit, we present that to prune the tree.4、According to the existing problems that the facial expression information will be lost under partial occlusion,We add the experiment of facial expression recognition under occlusion, the typical occlusion form such as eyes, mouth, the left part and the right part.