Research On Realization And Application Of Neural Networks On FPGA

The imitation and simulation of neuron and neural networks are the cores of brain-inspired computing.Field programmable gate array(FPGA)has the advantages of parallel computing,flexible programming and high integration,which will be very suitable for large-scale neural network simulation.The central pattern generator(CPG)is the basis unit for generating rhythmic movement in the motor nervous system.This paper mainly discusses the realization of the neural network on FPGA platform and study the mechanism of the locomotion neuron network based on this platform.First of all,we presented the design framework of FPGA-based neural network simulation system,and systematically discussed about the hardware design,software user interface and data communication of the system respectively.Meanwhile,a good extension scheme for large-scale neural network hardware simulation is given.We proposed a multi-electrode arrays(MEA)simulation platform based on the FPGA,and designed a software user interface based on LabVIEW,which can provide an easy,intuitive and convenient high-speed simulation platform for scientific personnel.Secondly,the dynamic characteristics and rhythm generation mechanism of central mode generator(CPG)were deeply investigated in order to understand the biological motor nerve system.We realized the minimum CPG neural network on the FPGA platform which meets the demand for real-time simulation of neural networks.It's helpful to reveal the mechanism of rhythmic movement generation and modulation in CPG networks and has important significance on bio-inspired robots design and locomotion recovery treatment.Besides,the use of the optimized multiplier-less hardware implementation method greatly reduces the hardware resource consumption,which lays the foundation for the hardware realization of large-scale neural networks in the future.Finally,the respiratory CPG neural network is successfully implemented on the FPGA simulation platform,and the corresponding dynamic characteristics and rhythm control mechanism of the respiratory CPG neural network are also analyzed.In addition,this paper proposes and designs a FPGA synchronous control simulation platform of the cardio-pulmonary system based on CPG networks,which will has animportant influence on the treatment of respiratory diseases and respiratory rhythm regulation.The high-speed FPGA simulation platform will also provide new ideas for the brain-inspired computing and the simulation of neural networks.