Research On Algorithms Of Face Recognition Based On Neural Network

As one of the most important applications of image processing and pattern recognition, research in face recognition have been increased significantly in many fields.Many subjects such as image processing, pattern recognition and computer vision are included in the problems of pattern recognition. There are two parts of the face recognition system: feature extraction and pattern recognition. Feature extraction is to extract useful information for classification from face images; while pattern recognition is to classify the features that extracted.For the feature extraction part, the method used in this thesis is based on statistical features. They are eigenface method based on KL Transform or Principal Component Analysis (PCA) and Fisherface method based on Linear Discriminant Analysis (LDA).For the pattern recognition part, the methods used are the SVM's with error correction and the Hyper-Ellipsoid Neural Network (HENN). The SVM's with error correction method has a high correct recognition rate; while the HENN method is one of the applications of biomimetic (topological) pattern recognition theory, which has a very low false recognition rate. Based one the two points mentioned above, this paper presents a two-pass classification method.It combines the HENN method based on the biomimetic pattern recognition theory and the SVM method with error correction ability. The HENN is used for the first classification to get the intermediate result, and the SVM with error correction method is used to solve the intermediate result and all the training samples for the second classification. It has the advantages of both biomimetic pattern recognition method and the SVM method with error correction ability, the first is that it can avoid the high false recognition rate, and the second is that is has the error correction ability, so it outperforms either of the two methods. Simulink experiments on Cambridge ORL database shows that with the Fisherface features the two-pass classification method, the correct recognition rate has reached 99.25% for image recognition.