Simulation And Hardware Implementation Of The Minitype Neural Network Based On FPGA

Neuron is the basic unit of the neural system,its main function is to receive,sort out and to convey neural information.Synapse is the important structure of information transfer between neurons.Neural network is set as the reasearch object which is made up of neuraons and chemical synapses.It is to model and to implement the neural network using the Hodgkin-Huxley model with biological neural electrophysiological properties as neural mathematical model and the excited Rabinovich model as synaptic mathematical model.Specific content as follows:(1)Basic konwladge of the neural network and its mathmatical model.The basic konwledge of neurons and synapses is introduced in this paper.The relationship between the neural circuits and the minitype neural network is elaborated in detail.It is illustrated the production and tansmission of neural electrical signals and the classification of the potential on the both sides of the neuron membrane.It is generalized mainly neural model and synaptic model.(2)The simulation research of the neural network.It is simulated divergence neural network,polymerization neural network,chain-type neural network and ring-type neural network based on HH neuron model and Rabinovich synapse model.The influence of the synaptic coupling strength on the divergence transmission of the neural electrical signals is studied.The influence of the polymerization on the neural action potential is evaluted.It is expiored the transmission mechanism of neuranl electrical signals in neural network by using different current to stimulate the chain-type neural network and the ring-type neural network.It is validated the consistency of DSP Builder software simulation with Simulink software simulation.(3)Hardware implementation of neural network.The chain-type neural network and the ring-type neural network are implemented which are made up with HH neuron model and Rabinovich synapse model by using FPGA hardware.According to the simulation of neural network,chain-type neural network and ring-type neural network are achieved by using QUARTUS ? software combination of DSP Builder techonlogy.In order to get the results of hardware,the neural network is applied different current stmulus.It is verfying the hardware of neural network with biological neural electrophysiological properties is right by comparing the results of hardware with the results of simulation.