A real-time silicon cerebellum spiking neural model based on FPGA

Abstract

Sensorimotor control and learning require the function of sophisticated neural system. Cerebellum is one such brain region which comprises more than half of the total neuron population in the entire brain. Capable of simulating a bio-realistic cerebellum model provides important information for neuroscience and engineering. Here we present a Network-on-Chip (NoC) hardware architecture for implementing a bio-realistic cerebellum model of passage-of-time (POT) encoding with 100,000 neurons. The results demonstrate that our implementation can reproduce the POT functionality properly. The maximum computational speed can reach 25.6 ms for simulating 1 sec real world activities. Our silicon cerebellum can be readily interface with in vivo or in vitro experiment and be adapted as a potential neuroprosthetic platform for future biological or clinical applications.