Application of flexible neuronal matching for face recognition

Abstract

In this paper, a model-based face recognition approach using a pyramidal Gabor function representation and neural network matching system is proposed. The local feature extraction is base on distortion tolerant Gabor transformation. The system is based on the Dynamic Link Matching. It consists of an image layer and a model layer, which are tentatively simulating primary visual cortex and infero-temporal cortex. It is inherently invariant with respect to shift, and is robust against many other variations, most notably rotation in depth and deformation. The system requires very little genetic or learned structure, relying essentially on the rules of rapid synaptic plasticity and a priori constraint of preservation of topography to identify matches.