Multi-FPGA Implementation Of Basal Ganglia And Parkinson's State Analysis

With the rapid development of computational neuroscience,the use of computers to realize the biological characteristics of neural networks has become a hot spot for neuroscientists in recent years.However,the characteristics of the computer's serial operation make it essentially different from the human brain.This difference has become more and more obvious with the increase of the neural network model,and the parallel computing ability of FPGA(Field Programable Gate Array)is a new solution to solve these problems.Therefore,it is important to propose that to implemente the basal ganglia and analysis the Parkinson's state based on multi-FPGA.Firstly,the structure and function of the multi-FPGA hardware simulation platform are studied.The problem of clock synchronization between FPGA chips in multi-FPGA hardware simulation platform is analyzed.The idea of using multi-chip FPGA to realize the basal ganglia neural network is put forward,which solves the problem of complex neural network is difficult to achieve using a piece of FPGA in current research.Secondly,the biological discharge characteristics of FHN(FitzHugh-Nagumo)neural network are realized.The FHN neural network model is built by the tool of DSP Builder,and the discharge characteristics of FHN neuron network are analyzed on the multi-FPGA simulation platform.The effect of synaptic currents on the FHN neuron network is summarized,which lays the foundation for the study of the implemention of basal ganglia.Then,the biologic dynamic characteristics of the basal ganglia are realized on the multi-FPGA simulation platform.The mathematical model of Izhikevich neuron is optimized by the piecewise linear approximation method.The optimized Izhikevich neuron mathematical model greatly reduces the consumption of the logical resources compared with the original Izhikevich neuron mathematical model.The paper introduces the process and method of using Izhikevich neural network to build the basal ganglia network,and have a software simulation on it.Download up the basal ganglia network to the multi-FPGA simulation platform for hardware simulation,and implemente the biological dynamics of the basal ganglia in multi-FPGA simulation platform.Finally,the discharge rhythm of basal ganglia neural network is studied.The discharge rhythms of GPe,GPi,STN and TC are studied on the multi-FPGA simulation platform.The comparison between hardware simulation results and software simulation results shows that the multi-FPGA simulation platform can fully realize the dynamic characteristics of large scale neural network such as basal ganglia,which lays the foundation for the design of a hardware simulation platform suitable for various large scale and complex neural networks in the future.The results show that compared with the original Izhikevich neuron mathematical model,the Izhikevich neuron mathematical model optimized by piecewise linear approximation method can greatly reduce the consumption of hardware resources,which provides an important reference for the larger scale neural network simulation in the future.In addition,the multi-FPGA simulation platform is used to realize the biological dynamics characteristics of large scale neural networks such as basal nucleus,which provides a reference for the future design of a hardware simulation platform that can be applied to all kinds of complex neural networks.