Hmm-based Face Recognition Methods

Hidden Markov Model (HMM) is a statistical method based on the overall probability, it features linked to various organs numerical description of the human face, thus achieve very good results.This paper presents an anisotropic diffusion method for facial image preprocessing based on wavelet transform. Firstly, the image is decomposed by discrete wavelet transform. Secondly, every vector is denoised by means of anisotropic diffusion method. Comparison and experimental result show the new method has high performance.In this paper, a new algorithm for face recognition is proposed, based on wavelet transform and hidden Markov model (HMM). A sequence of overlapping sub-images is extracted from each face image, computing the wavelet coefficients for each of them, and the low dimension subspace was formed by using PCA analysis. And the whole sequence is then modeled by using Hidden Markov Model. Experimental results show that the proposed algorithm has a high recognition rate in different illumination condition.In this paper, we made further improvements to the MMD algorithm. And Maximum Model Distance (MMD) is used on E-HMM training. Under the new algorithm, we also got the estimation equations; we gave the simple matrix form of the parameters estimation. Experiments show that this algorithm could be used to better effect.