Spiking Versus Traditional Neural Networks for Character Recognition on FPGA Platform

Abstract

Spiking Neural Networks (SNN) is considered the third generation of neural networks. This type of neural networks are inspired from biological nature of cortical neuron, and they (SNNs) introduced the concept of time rather than using real-valued inputs and outputs, which is a characteristic feature of Artificial Neural Network (ANN). The purpose of this research work is to design, develop and implementation of character recognition algorithm in a hardware environment (FPGA) based on Artificial Neural Networks and Spiking Neural Networks. Then, a comparison study is made, between these two generations of neural networks, regarding execution time and hardware size on the FPGA platform. Altera CycloneII DE2 (development and education) board has been suggested to work with for design and implementation of character learning and recognition. Coding with Verilog Language has been used for developing and synthesising the software structure of SNN and, also, to perform learning and recognition inside Altera Cyclone-II DE2 hardware. The learning method used in ANN is back propagation, while spike time dependent plasticity (STDP) has been employed for learning of SNN.