Hidden Markov Model And Neural Network Face Recognition Algorithm

Automatic Human Face Recognition (AHFR) is a leading research field in Computer Science; AHFR could be used in many disciplines. To employ machine in AHFR, we need to pay attention to these fields: pattern recognition, digital image processing, cognitive science and psychology, etc. There are some popular methods for AHFR: eigenfaces, wavelet analysis, neural networks based recognition and statistical model based recognition. But we may encounter some problems, e.g., there will be low recognition efficiency when the illumination, angle and the size of human face varying, and there is another problem when we use neural networks to recognize human face, the original gray images contain a hugeness data, these data requires tremendous neurons and this is a time consuming task.To overcome the problems listed above, in this thesis, I give two novel algorithms for AHFR: one is Hidden Markov Model (HMM) based method and another is mathematical morphology based bp neural networks.The contents of this thesis are arranged as follows:1. First I analysis the state of the art of the researches of the AHFR; then I give a discussion on the critical techniques such as extraction of the characteristic of face, HMM and the reconstruction of neural networks.2. I give a brief solution on the building of such a AHFR system and then I show the system architecture.3. Because that for a human face, the significant facial regions from top to bottom and the local characteristic regions from left to right can be described as a state sequences of HMM, so here I give a composite method which incorporates 2D-DCT based data compress method and HMM. This method performs fast human face recognition by computes the most similarity of the human face according to the optimistic state sequences.4. I show my researches on the Mathematical Morphology based characteristic extraction of human face; the experiment shows a good result. 5. I discuss on the 3D reconstruction of human face model.6. Researching on using COM in the Visual C++ and Matlab mix programming, which leads to a fast system development.