Research On Face Recognition Technology Based On Relevance Vector Machine

With the rapid development of information technology, face recognitiontechnology has become the important part in the field of pattern recognition, andplays an important role in identity authentication. Face recognition has theadvantage of being more secure, reliable and efficient, and has been applied tomany fields.Face recognition is mainly composed of four parts: face detection, imagepreprocessing, feature extraction, classification and recognition. The images usedin this article are images that contain the face only, so face detection is notneeded. With regard to image preprocessing, scale normalization and graynormalization is adopted in this dissertation. In feature extraction processing,wavelet was firstly used to extract the principal information of face. Then, globalfeatures of the face images were extracted by using PCA method. Featureextraction effectively extracted the main features of faces and reduced thedimension of images, and part of the interference information was filtered. Inclassification and recognition processing stage, a multi-classifier based on RVMwas presented in this dissertation, and the1-v-1multi-classification method wasimproved.Relevance Vector Machine is a machine learning algorithm which was firstproposed by Micnacl E. Tipping in2000. This method is based on Bayesianestimation theory. It is suitable for processing regression and classificationproblems, and has the good generalization performance and generalization ability.Since the birth of RVM, RVM had been paid more attention by many researchers,and more RVM relatived researchs are being studied.With the simulation experiments, based on RVM algorithm face recognitioncan get higher recognition rate than SVM algorithm was verified by applying ORL face database in this dissertation. As shown in the simuliaiotn result, theface recognition appling self-built face database based on RVM is also suitablefor Asian faces. With the using of appling the improved RVM multi-classifier onmixed face database it shows the improved classification algorithm can improvethe speed of face recognition.