Face recognition based Hybrid Fuzzy Hidden Markov Models

This paper proposes Hybrid Fuzzy Hidden Markov Models (FHMM) for face recognition. This recognition system includes fuzzy integral theory and Hidden Markov Model. Applying fuzzy expectation-maximization (FEM) algorithm in the Hidden Markov Model (HMM) is to estimate the relative parameters of faces which are close to real values in a better condition. Besides, in order to precisely obtain the probability density function of observations vector, taking full use of Gaussian Mixture Models (GMM), in which the weights are designed by using the fuzzy c-means (FCM) function. Comparing to conventional HMM, the proposed method achieves a better result.