抄録
We developed a new algorithm for face recognition using a Bayesian framework. In our algorithm, posterior distributions are computed using a Hamiltonian Monte Carlo method. The face features we used were those used in our previous work based on the elastic graph matching method. While our previous method attempted to optimize facial feature point positions to maximize a similarity function between each model and face region in the input sequence, our new approach evaluates posterior distributions of models conditioned on the input sequence. Experimental results show that our new algorithm out-performs the Eigenface method in terms of both identification errors and processing costs. Our new algorithm eliminates almost all identification errors on sequences showing individuals smiling, although such data was not used in training.
