Face Recognition Using Foveal Vision: Preliminary Report

Data from human subjects recorded by an eyetracker while they are learning new faces suggests face recognition can be modeled as a sequential stochastic process, where the underlying observations at a given fixation depend on both foveal and parafoveal information. In contrast to most pattern recognition based approaches, this foveal face recognition approach is incremental and scalable to large face images. This paper describes two approaches to implementing an artificial fovea, which transform a constant resolution image into a variable resolution image with acute resolution in the fovea, and an exponential decrease in resolution towards the periphery. For each individual in a database of faces, a hidden-Markov model(HMM) classifier is learned, where the observation sequences necessary to learn the HMMs are generated from foveated images, by fixating on different regions of a face. Detailed experimental results shows the two foveal HMM classifiers outperform a more traditional HMM classifier built by moving a horizontal window from top to bottom on a highly subsampled face image.