Face Recogntion Algorithm Based On Gabor Feature

Human Face Recognition is an important branch of recognition pattern, it becomesthe focus in biological features recognition with intuition and noncontact. In recent years,domestic and international scholars study and find many mature theories and algorithms.People can locate and recognize the different faces, but some limitations still exist incomplicated conditions. Meanwhile, because the limitations of the computer performanceand storage capacity, many perfect algorithms have deficiencies in recognition time andRecognition Accuracy.This paper studies algorithms for facial features extraction, it proposes a newalgorithm. It also presents a new dimensional reduction against over-high dimensions andconsiderable computational complexity. The new algorithm is helpful to improverecognition accuracy by the experiments, it has the certain theory and the practicalvalue.This paper mainly studies the main following aspects:1.The paper emphatically analyzes algorithm extraction advantage and disadvantagebased on wavelet feature, and proposes an algorithm based on discrete waveletdecomposition, reconstructed picture and weight value wavelet coefficient.The.differential recognition power theory is introduced to algorithm extraction based onGabor wavelet feature. It introduces weighted-analysis method after two-dimensionalwavelet discrete analysis images. Then higher weight value of the face features areselected, the result is feature dimensions decreased and the identification accuracyimproved.2.Gabor wavelets have excellent locality and selectivity, according with the humanvision recognition. Therefore, it becomes a vital method of face recognition. With thegood recognition rate, the higher dimensions take some problems in calculation conditions.After studying the principal component analysis algorithm thoughts and the applicationof second decimation algorithm about PCA and2DPCA, this paper gives a newclassification and dimensional method which is different from wavelet kernel, it providessome constructive suggestions about integrating Gabor wavelets and2DPCA by thepremise of improving calculation efficiency.3.This paper researches the classifers for the part of face recognition. The radial basisfunction is chosen as the kernel of support vector machine, The algorithms proposed inthis paper are proved by experiments and compares with other algorithms. Theexperiments results.show that the algorithms proposed in this paper are effective forimproving the face recognition accuracy.