Multiplierless Implementation of Fitz-Hugh Nagumo (FHN) Modeling Using CORDIC Approach

The study, simulation, and implementation of neural behavior in the human brain are central goals of neuromorphic engineering. By integrating various scientific fields, we present a hardware solution based on neuronal cell mechanisms that can emulate such a nature-inspired system. This article presents a Fitz-Hugh Nagumo (FHN) neuron implemented using COordinate Rotation DIgital Computer (CORDIC), which accurately reproduces various patterns of the original FHN neuron model. We propose a modification to the original nonlinear term using a CORDIC IP-Core, resulting in high matching accuracy and low computational error. The proposed model is validated through time domain and dynamic analysis, which demonstrates its high accuracy and low error in reproducing all features of the FHN model. For large scale neuron implementations, we present an efficient digital hardware solution based on the resource sharing techniques. The hardware is implemented on Field-Programmable Gate Array (FPGA) using Hardware Description Language (HDL), as a proof of concept. The results from the hardware implementation show that the proposed model uses only 1% of the resources available on a Virtex 4 FPGA board. Additionally, the static timing analysis shows that the circuit can operate at a maximum frequency of 320 MHz.

I. Introduction

Spiking Investigating the Central Nervous System (CNS) requires a deeper understanding of Neural Networks (SNNs), as they make up the fundamental building blocks of the CNS, consisting of neurons and synapses [1], [2], [3], [4]. The study of SNNs has numerous practical applications, including pattern recognition, data processing, medical diagnoses, and autonomous robotics. The CNS is an intricate network of neurons, synapses, and the calcium-based cells called glia, which plays a critical role in protecting and regulating the behavior of neurons and synaptic coupling [5], [6]. To gain a comprehensive understanding of the brain's neural network, it is imperative to study the activity of neurons through mathematical modeling. Thus, neural modeling and its mechanisms have become crucial tools for analyzing and processing the behavior of biological neural networks within the CNS.