Ehmm-based Face Recognition

Face Recognition is that recognizing identity which is belong to the face from one or some of static or video image sequence. The sequence is gained from be known identity face information data base. In generally speaking, Face recognition system is made up of five groups ,which is Face Detection, Face location and Tracing, Face Representation, Face Recognition, Expression/Gesture analysis and Face physical classification. In the face recognition system, face detection and face recognition are the most important facets. They also determine the face recognition system's correction and efficiency. In my thesis, I exploited 2D-DCT in face eigenvalue extraction. In the transform, one the one hand it make the picture's dimension from high to low, one the another hand, after the transform from space to frequency, choosing some low frequency coefficient as the face eigenvalue, so it will improving the later face training and face recognition's robusticity. Be aim at the little swatch's training question. In my thesis, every face object have limited swatch, t is very important how use the limited swatch sufficiently to train face object and In order to get the trained face mode with good robusticity as soon as possible. So in my thesis take multiswatch weight synthetic trained method .As the information is gained from the different face picture ,so will take different weight in it .After trained, use one synthetical mode figure out one face mode. From the result of face recognition which the face is switched from face base have been collected. It is obvious that the method have gained better recognition rate and rapidity.