基于FPGA的小型神经元网络的模拟与实现

发布时间：2018-02-24 18:34

  本文关键词： 生物神经元网络 Hodgkin-Huxley模型 化学突触 数值模拟 现场可编程门阵列　出处：《兰州交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：神经元是构成神经系统的基本单元,其主要功能是接收、整理和传递神经信息;突触是实现神经元与神经元之间信息传递的重要结构。神经冲动的传导与传递是研究神经系统功能的重要方面。本文主要以神经元与化学突触构成的神经元网络为研究对象。采用具有生物神经元电生理特性的Hodkgin-Huxley模型作为神经元数学模型,采用兴奋的Rabinovich模型作为突触得数学模型,对规则的神经元链网络与神经元环网络进行仿真模拟和硬件实现。具体包含如下内容:(1)神经元网络基础及其数学模型。主要介绍神经元与突触相关的基础知识。对神经回路与小型神经元网络的关系进行详细阐述。对神经电信号在神经元网络中的产生与传播以及神经元细胞膜两侧电位的分类进行说明。概括现在建模主要采用的神经元模型以及突触模型。(2)神经元网络的仿真模拟研究。对HH神经元模型与Rabinovich突触模型组成的辐散神经元网络、聚合神经元网络、神经元链网络与神经元环网络进行仿真模拟。研究突触耦合强度对神经电信号的辐散传播的影响,聚合作用对神经元动作电位的影响;采用不同的刺激电流对神经元链网络和神经元环网络进行刺激,探索在不同刺激下,神经电信号在神经元网络中的传播机制,并对DSP Builder与simulink软件仿真的一致性进行验证。(3)神经元网络的硬件实现。运用FPGA(Field Programmable Gate Array,现场可编程门阵列)对HH神经元模型与Rabinovich突触模型组成的神经元链网络与神经元环网络进行硬件实现。依据仿真描述的神经元网络,运用QUARTUSⅡ软件结合DSP Builder技术,完成神经元链网络与神经元环网络的FPGA硬件实现。对神经元网络硬件施加不同的电流刺激,得到神经元网络的硬件实现结果。对比神经元网络的硬件实现结果与仿真模拟结果,对硬件实现具有生物神经元电生理特性的神经元网络的正确性进行验证。
[Abstract]:Neuron is the basic unit of nervous system, whose main function is to receive, organize and transmit neural information. Synapse is an important structure for the transmission of information between neurons and neurons. The conduction and transmission of nerve impulses is an important aspect of studying the function of nervous system. In this paper, the neuronal network composed of neurons and chemical synapses is mainly used. The Hodkgin-Huxley model with biological neuron electrophysiological characteristics was used as the mathematical model of neurons. Using the excited Rabinovich model as the synaptic mathematical model, The simulation and hardware implementation of regular neuronal chain network and neuronal ring network are carried out, including the following contents: 1) the basic and mathematical model of neuron network. The basic knowledge of neuron and synaptic connection is mainly introduced. The relationship between neural circuits and small neural networks is described in detail. The generation and propagation of nerve signals in neural networks and the classification of neuronal cell membrane potentials are explained. The main modeling methods are summarized. The neuronal model of HH and the synaptic model of Rabinovich were used to simulate and simulate the neural network. The divergence neuron network composed of HH neuron model and Rabinovich synaptic model was studied. The effect of synaptic coupling intensity on the divergence of nerve signal and the effect of aggregation on action potential of neurons were studied. The neuronal chain network and the neuronal loop network were stimulated by different stimulation currents to explore the transmission mechanism of the nerve signal in the neuron network under different stimuli. The consistency of DSP Builder and simulink software simulation is verified. The hardware implementation of the neuronal network is verified. Using FPGA(Field Programmable Gate array, the neuronal chain network and neural network of HH neuron model and Rabinovich synaptic model are made up of HH neuron model and Rabinovich synaptic model by using FPGA(Field Programmable Gate array (FieldProgrammable Gate Array). According to the simulation description of the neural network, Using QUARTUS 鈪，

本文编号：1531291