FPGA implementation of an excitatory and inhibitory connectionist model for motion perception
Abstract
Based on neurophysiological studies of the visual cortex, the design and FPGA implementation of a bio-inspired hardware architecture for visual perception of motion is presented. The architecture is based on a neural model that mimics the way in which the visual stimuli is processed in the visual cortex pathways by means of spatio-temporal filtering and excitatory-inhibitory neural processing. This paper mostly develops the time-multiplexed and connection-parallelized implementation of the densely interconnected module that performs excitatory and inhibitory computation loops. The architecture modules have been modeled in VHDL and independently synthesized for an FPGA device. The obtained results show the plausibility of efficient implementation of bio-inspired models through embeddable specialized hardware architectures restricted to real-time constraints.