Research On Closed Loop Modulation Method Of Neural Oscillation

It is a trend for brain science study around the world in this century.Continuing and substantial brain science projects have been commenced in the EU,US,Japan and China,among others.Research on brain science shows that cognitive,behavioral and perceptual brain functions originate from the collective nerve oscillations in the cerebral cortical system,and the collective nonlinear dynamics is the core of the cortical nerve oscillations,and the abnormal dynamic processes will cause many brain diseases.Therefore,it is crucial to implement theoretical studies related to the modulation of neural oscillations,which can furnish scientific evidence for the comprehending of the neural process of human cognitive function and the pathologic mechanism of brain diseases,and supply scientific support and technical guarantee for the enhancement of human cognitive function and the development of effective treatment methods for brain diseases.This paper combines control theory and neuroscience,based on the high-order nonlinear neural mass model proposed by Jansen and Rit to simulate the neural oscillation of the brain,to study the design of the closed-loop feedback modulation method.Firstly,a Mamdani fuzzy neural oscillation closed-loop modulation method based on genetic algorithm is designed for highly nonlinear neural group model.The design of this modulation method does not bank on the accurate mathematical model of the plant,and it is notably appropriate for nonlinear systems,and the troublesome problem,like the development fuzzy rule and membership function,is dealt with the improved genetic algorithm.Secondly,to settle the shortcoming that the output magnitude of the fuzzy modulation method is too large,an adaptive neural oscillation closed-loop modulation method is designed by using the backstepping method.This modulation method can recoup for the unknown average excitatory synaptic gain of plant,has fewer adjustable parameters and demonstrates the asymptotic stability of the closed-loop system with the help of the Lyapunov function.Thirdly,to settle the shortcoming that the output magnitude of the fuzzy modulation method is too large and the adaptive modulation method banks on plant,a predictive modulation method is designed.The prediction model of this modulation method is solved by the radial basis neural network,which is used to recognize the input and output of the neural oscillation model in real time by the gradient descent method.It has strong robustness and the output amplitude of the modulation is constrained by the solver.Finally,the feasibleness and validity of various closed-loop modulation schemes are confirmed by Simulink.At the same time,C code generated by Simulink simulation framework is deployed to Raspberry Pi,and the simulation results are compared in different platform.