Face Recognition Algorithm Based On LBP And Fisher Face

In biometric recognition-based identification, research on face recognition startsearly and its technology is relatively mature that it is widely used in many computerintelligent recognition fields. Facial feature extraction and expression is a key part offace recognition that it is closely related to the effect of face recognition. On the onehand, different feature expressing methods and dimensions of feature vector both willaffect the recognition rate. Under normal conditions, if the same feature expressingmethod is used, the higher the dimension of feature vector, the higher the recognitionrate. On the other hand, value of dimension of feature vector extracted from facialimage will directly influence the computing complexity of face recognition, the largerthe dimension of feature vector extracted, the more running time it needs to recognizethe algorithm.This article focuses on facial feature expression and recognition method whichutilizes local pattern descriptor, discusses using Fisher face method to reducedimension towards feature vector and improve its recognition effect at the same time,thus to meet the actual needs of face recognition system. The main work of this paperembodies in the following two aspects:(1) It conducts in-depth study on facial feature extraction methods based onPrincipal Component Analysis (PCA), Fisher Linear Discriminant Analysis (FLD),local binary pattern and local ternary pattern. Compared with global pattern descriptor,the effect of face recognition is more ideal if local pattern facial feature expression isput to use for it greatly improves the recognition rate of face recognition. Comparedwith local binary pattern descriptor, the anti-noise ability of descriptor based on localternary pattern is stronger.(2) For feature vector, Fisher face method not only can effectively reducedimension towards feature vector extracted, but also can get good effect in lineardiscriminant, a facial feature extraction method which combines Fisher face methodand local pattern descriptor is obtained thereby. First reduce dimension towards facialfeatures extracted by making use of local pattern descriptor through introducing Fisher face method, then discriminate and analyze the feature vector after dimensionreduction, thus improving the speed of face recognition and improving the recognitionrate at the same time.