Research Of Face Recognition Algorithm Based On Fuzzy Rules

The Face Recognition Technology covers wide fields of human brain cognitive science as well as subjects of image processing,pattern recognition,artificial intelligence and signal processing.This is a challenging research topic. With the development of computer technology and applications, face recognition also has been applied in many fields.An important step to identify is the feature vector extraction, although the existing variety of feature extraction methods, but the Local Binary Pattern (LBP) is my choice to extract the feature vector.LBP is a very powerful method to describe the texture and shape of a digital image. Therefore it appeared to be suitable for feature extraction in face recognition systems.Different part of the face for rencognition is different,so in this paper we proposed a method-Face recognition algorithm based on fuzzy rules.First encoding the entire image with LBP,and then divided the encoded image,followed calculate the variance and entropy of each sub-block region and design an fuzzy controller,the fuzzy controller’s inputs are variance and entropy.thc output is weight.Finally, the weights assigned to the corresponding sub-block area and connected in scries to each sub-block region vector obtained eigenvectors.Experiments on the face database, shows the weighting algorithm more better.This paper also discusses the extension of the recognition algorithm, and finally found through experiments, the expansion of the neighborhood can not improve the recognition performance, it increases the length of the feature vector and the recognition speed become slowly, and the recognition rate lower Accordingly, so this attempt can be discarded. Another extension is LBP combined with PCA, PCA is used to reduce the dimension, This can save calculation time and possibly improve the recognition rates. A possibility is to apply PCA on the original feature vectors of the different regions. In this case the weights still can be used during calculating the distance matrix.so this is a good extension.