The Study Of Face Recognition Method Based On Wavelet Transform

As one of person’s unique physical characteristics facial features has become a major virtue by wich to distinguish different individuals.Thanks to the advantage of non-contact, Face Recognition has been well accepted by the public. During recent years, with the development of technology, especially the development of computer technology, providing the necessary conditions for the improvement of face recognition technology.This paper presents a combination of principal component analysis (PCA) and Fisher linear discriminant (FLD) based on wavelet transform, and nearest neighbor method to identify different faces.The action of wavelet transform is retaining key information, removing the noise and redundant information, reducing the quantity of mathematical operation. In this paper, firstly decomposing the faces by the wavelet transform, then choosing low-frequency components obtained from the wavelet decomposition to reshape the faces. The purpose of this step is to remove noise and redundant information. Secondly extracting the features with the PCA+FLD method. The use of the PCA method is achieving a lower dimension, so that we can improve the efficiency of the face recognition.While the use of the FLD method is projecting the low-dimensional vector onto a direction in wich can separate photos of different testers easily,so that we can improve the accuracy of face identification. Finally, using the nearest neighbor method to classify the faces and out putting the result. In order to compare the different effects on the accuracy of face identification by using different wavelet, different levels of decomposition and different photo reshaping components. Corresponding cods has been compiled under the MATLAB environment to calculating the accuracy of face identification and the time spend on face identification.By comparing the experimental data.A effective face recognition system with high accuracy level has been design and realized.