Method Based On Wavelet Multiresolution Analysis And Kernel For Face Recognition

Face recognition is one of main methods of status authentication which basedon technology of biologic characters recognition. Automatic status authenticationbased on person face recognition has great theory significance and application value.It attracted people's great concerns since 1960s and research on the methods ofperson face recognition has become hot topic in the field of pattern recognition andartificial intelligence. Recent years the technology of face recognition hasexperienced a rapid development, especially, urged by computer technology,recognition by computer has become a very important new-born recognitiontechnology. As non-rigid body, person faces are characteristic of frequent shapechange, too much influence factor and it was also easy to be disturbed. Besides,owing to the confine of condition, it was impossible to gather enough samples. Socompared with graphic vector, it was a problem of small samples. Many traditionalpattern recognition methods tend to be over-study and under-study on such kind ofmulti-dimensional and nonlinear problem.This paper firstly introduced the background and main methods of thetechnology of person face recognition with computer. It analyzed the difficulties andphysiological essence of person face recognition; summarized the main methods ofperson recognition; put forward a person face recognition method which based onGabor wavelet and Fisher analysis. That is to say, before the nonlinear mapping, itfirstly used wavelet transformation to preprocess the original person face graphs,then picked up the characters from the graph of person face by Gabor wavelet andmerged all those characters, at last it did core-Fisher analysis on the range offrequency. According to the experiments, the combine of these two methods hasreceived remarkable recognition rate on person face recognition. At last, wesummarized and look forward to the further study.