The Analysis Of Response Characteristics Of Local Field Potential Based On Adaptive Filter

To the brain information processing mechanism and brain-machine interfaceresearch，it is very significant that extract the neuron response signals related to thestimulus from neural signals. However, due to acquire the neuron signals frommicroelectrode array of extracellular mode, signals are small value, vulnerable to theinterference of noise and distorted, especially the local field potential signals arestrong non-stationary, non-Gaussian and nonlinear, which make extract the neuronresponse signal from the local field potential with great challenge. Therefore,researching noise suppression methods of the local field potential signals andimproving SNR of the signals become one of the key and core problems based onthe neural information processing of the local field potential.In this thesis, based on the deep analysis of the local field potential and thestatistical properties, we mainly study adaptive algorithm to extract response signalrelated to the stimulus neuron signal of the local field potential, then analyze theanimal neuron response characteristics of the typical condition. The main results areas follows:1: We firstly analyze the mechanism of the generation of the local fieldpotential, show the statistical features of the local field potential signals, the resultsindicate that for different channels, the local field potential waveforms collectedfrom microelectrode array of extracellular technology are of high similarity, and thesignal energy of spontaneous state is a downward trend with the increasing frequency,signal energy is mainly concentrated on the low frequency, and signals show thetypical non-stationary characteristics.2: We analyze the noise types and their sources of neural signals, on the basisvirtual reference technology will be introduced to denoise the local field potentialsignals, and analyze the feasibility and existing problems extracted neuron responsesignal with the difference average reference technology and common technology inthe local field potential, and the results show that it is feasible using the virtual reference technology to denoise the local field potential signal, and compared withdifferential reference technology, common average reference technology is moreeffective.3: In view of the problems that exist in the denoising local field potentialsignals with the virtual reference technology, we combine the adaptive filter theorywith the virtual reference technology, and put forward four design methods ofreference channels of the adaptive filter, and verify the four design methods with themeasured data. Results show that no matter from time domain or frequency domain,the four algorithms can improve the signal-to-noise ratio of the local field potentialsignals, however, in comparison, denoising effect based on weighted averagereference technology combined with the adaptive filter is best.4: We analyze the three typical states of rats, namely response characteristics ofthe local field potential with the silent gaze, exploration, chewing. Aimed at theresponse characteristics of three typical states in the local field potential, we presenttwo kinds of response features of neurons with significant differences, and use thetwo kinds of neuron response features, based on Bayesian estimation theory, todecode and analyze the activity of rats, it conducts a meaningful attempt for theresearch of the brain information processing mechanism.