The Research Of Face Feature Extraction And Recognition Based On Modular

Face recognition technology is a very active problem in the field of computer vision and pattern recognition. Feature extraction is the most important in the area of pattern recognition. The key of the face recognition technology research is how to extract discriminate in favor of classification features. In this paper, face image change is large by facial expressions, gestures and light conditions and the traditional feature extraction methods can not extract local features. At the same time, data dimensions are too high, fewer samples and identify problems is not ideal. Therefore, this paper is to do the following research based on the extracting feature theory and Modular method, the main work is divided into the following sections:(1) The background, significance, contents and application of the face recognition are all discussed. We will give a brief introduction about face database in the last.(2) Detailing description face recognition algorithms of the current classical algebraic feature extraction method: Principal component analysis(PCA) and linear discriminant analysis(LDA) method. The advantages and disadvantages of these two methods were systematically explained.(3) This paper introduces a maximum margin criterion(MMC) algorithm to solve with the small sample size problem which is exists linear discriminant analysis method.The method use the difference of between-class and within-class scatter matrix as identification criterion, whether the within-class scatter matrix is singular or not that has no affect for final identification. However, MMC method has the following shortcomings: 1. MMC is easy to be influenced by the light condition and the variety of facial expression because it extracts the global feature of the image, and it can not extract information to identify strong local feature vector. 2. Need transform image matrixes into columns make high dimensions of the covariance matrix and a large amount of calculation. So we propose the modular MMC method. Modular MMC algorithm can partially eliminate the instability of face images caused by the different conditions of light to help extract effective local information.(4) 2DMMC direct method using image matrix scatter matrix structure avoids the small sample size problem. The covariance matrix dimension of the 2DMMC is farbelow of the MMC. Compared with MMC, it can greatly improve the speed and accuracy of feature extraction and reduce the complexity of it. Ultimately it improves the speed of recognition and the recognition rate. Even though 2DMMC method reduce the computing capacity, but it is not obvious. So, combining 2DMMC algorithm and Modular MMC projection algorithm, we proposed a new algorithm Modular 2DMMC.The dimension of each sub-block has a greater degree reduction after M2 DMMC method. The impact is not great even if they need to deal with much matrix. The dimension of the feature space has been greatly reduced; thereby reducing the system storage needs, the amount of computation is reduced accordingly.(5) At last, we use the MATLAB software with related proposed algorithm to create the face recognition system based on the theoretical in the previous section.