Neural encoding and spike generation for Spiking Neural Networks implemented in FPGA

Abstract

In this article a new digital system for generating spike pulses is presented. In real time, the system can convert digital values into artificial neural spikes for Spiking Neural Networks (SNN). The digital system can perform three basic functions: to generate spikes, to convert digital data values into pulse trains and, additionally, to encode spike trains on three major code classes reported in the scientific literature of neuroscience: rate coding, temporal coding or population coding. This system is therefore a digital interface between the physical world, computer systems, or digital data and SNN that are processed in real time. A functional prototype module was developed in a Ciclone IV FPGA using less then 300 logic blocks.