Feedforward Computational Model for Pattern Recognition with Spiking neurons

Abstract

Humans and primates are remarkably good at pattern recognition and outperform the best machine vision systems with respect to almost any measure. Building a computational model that emulates the architecture and information processing in biological neural systems has always been an attractive target. To build a computational model that closely follows the information processing and architecture of the visual cortex, in this paper, we have improved the latency-phase encoding to express the external stimuli in a more abstract manner. Moreover, inspired by recent findings in the biological neural system, including architecture, encoding, and learning theories, we have proposed a feedforward computational model of spiking neurons that emulates object recognition of the visual cortex for pattern recognition. Simulation results showed that the proposed computational model can perform pattern recognition task well. In addition, the success of this computational model suggests a plausible proof for feedforward architecture of pattern recognition in the visual cortex.