Research And Implementation Of Face Recognition Technology Based On Embedded Hidden Markov Model

Face is the most universal pattern in human vision, and therefore face recognition has become one of the most easily acceptable identification methods. In recent years, face recognition technology has been attached great importance, which has become one of the most successful application in computer vision, image analysis and understanding. In this paper, a large number of domestic and foreign literature materials on face recognition, published in recent years, is analysized, an in-depth research about face recognition technology is conducted, and structure, principle and achieve of the face recognition system based on embedded hidden Markov model are systematically analysized. The main work accomplished is listed as follows:Firstly, in the feature vector extraction stage, taking pixel values and eigenvectors as observation vectors are compared. Both two-dimensional discrete cosine transform and two-dimensional discrete wavelet transform as feature extraction method are focused on and compared through experiments.Secondly, in the model training and the recognition stage, in view of the complex structure of EHMM, and much time in training several samples which is hard to meet the real-time problem, a new training method which is based on multiple eigenvectors integration is proposed. Through experiment verification, this method is proved to be feasible, and can cut down sample training time effectively.Lastly, after the theoretical research and analysis, a face recognition system designed and implemented. There are three functional parts of the EHMM-based face recognition system: feature extraction, model parameters training and recognition. Every module of the system is clear in function, relatively independent, and in a low degree of coupling characteristics.