Certification, Based On The Identity Of The Video Content

The identity discrimiante technique, which is based on the living creature characteristic information such as the speech, iris and the DNA etc, has already walked into a pratical stage. But, current technique application all is built on higher hardware to deal with information. This paper mainly do research on the gate system from various method of the information data processing, which is used with household-use computers as a handle tool.The video information is widespread in the supervision system, and face characteristics are intrinsic and stable to people and strongly different from one to the others, they are also friendly and accepetable, thus they can be used as features for identity discriminate. Thus our resreach will be based on the facial pictures within the video frequency contents.Face recognition technique contains face detection, characteristic extraction, training and recognition. To satisfy real-time and accuracy as it is used in gate system, we analysis and compare the method that has exited in each research part, then put forward some improvements. We also propose two new methods in the training and recognizing part.In the part of human facial detection and fixed position, we presents an algorithm to face detection based on complex and dynamic vedio information. The method involves locating human face like regions by substracting two adjacent pictures of asreies and using skin color model to segement face. Color information is used to detect the skin area first, and the template is used to match the face area.Second, in the part of human feature extraction, we presents the theory of PCA algorithm, I-PCA algorithm, KPCA algorithm, FDA algorithm and their application to face recognition. Then, as the need of time and accuracy in gate system, we propose an algorithm as PCA/ I-PCA +KPCA+FDA, and the experim- ent proved its possibility.During the period of identity, the variety and renewal of the data are important. The back propagation (BP) algorithm was introduced to train the neural network for face recognition. As is known, BP algorithm usually needs too much iterative times and wastes time during the training. This algorithm combines the optimization of the characteristic vector and the adaptability of the neural network to improve the recognition rate and the robustness of the algorithm to noises. For sequence coming from the steep-descent method, we use extrapolation method for the sequence to accelerate convergence and...