Design And Implementation Of Image Reconstruction Software System Based On LFP Signal Of Pigeon Optic Tectum

At present,the research of animal robots mainly focuses on stimulating specific brain regions of animals and controlling animal behavior.The visual system is the main sensory system for animals to perceive the external environment information,and the visual perception information accounts for more than 80% of the perceptual information.Birds have developed visual system,using brain-computer interface technology to acquire visual system neuron signals and reconstruct external visual input is of great significance for animal robots to use the visual system for rescue reconnaissance tasks.This paper takes the pigeon as the research object,and designs and implements the image reconstruction software system based on the neuron response signal of the optic tectum(OT).The system takes image as visual input,uses multi-channel microelectrode array to collect the response signal of the OT neurons and separates the Local Filed Potential(LPP)signal,extract the amplitude-phase characteristics in the frequency domain of the LFP signal,and the reconstruction model between the image and the LFP signal is constructed,and finally the image reconstruction is realized.The main work done in this paper and the corresponding research results are as follows:(1)Designed an image reconstruction software systemThe characteristics of the pigeon vision system are analyzed,the design principle of the reconstruction system is summarized,and Combining the acquisition requirements of the OT region neuron signal and the subsequent signal processing process summarizes the design requirements of the system.According to the system design principle and design requirements,the overall design scheme of the image reconstruction software system is proposed.The scheme divides the reconstruction system into neuron signal acquisition analysis and image reconstruction model construction.Neuron signal acquisition and analysis mainly includes neuron signal acquisition and neuron response characteristics analysis.Image reconstruction model construction mainly includes image visual stimulus generation,local field potential signal feature extraction and reconstruction model construction.(2)Collected the neuron signal and analyzed the response characteristics of neuron.The neuron signal acquisition platform was built,and the corresponding equipment was selected according to the needs of the image reconstruction system,which prepare for the post-visual stimulation experiment to obtain the neuron signal.The checkerboard experiment with grey background black was designed,and the neuron response signal was collected.The location and size of the receptive field of the middle and shallow neurons in the OT region were determined by the firing rate characteristics of the Spike signal and the inverse correlation method.Four gray background black grating stimulation experiments in different directions were designed.The Spike time of the four-direction neuron response was extracted and the Raster plots was plots.It is found from the Raster plots that the Spike time of the neurons stimulated by the grating in different directions is basically consistent with the Spike trains,which show that the shallow neurons in the OT region are not selective to the direction.Prepare for subsequent image stimulation pattern design and neuron signal acquisition.(3)Constructed an image reconstruction modelA right-to-left scanning stimulus mode was designed to solve the problem that the microelectrode array could not cover the whole image effectively.The neuron signal after image stimulation is obtained.The LFP signal was obtained from the neuron signal by low-pass filter,and The LFP signal was preprocessed by independent component analysis.The amplitude-phase characteristics of LFP signal in frequency domain was extracted by short time Fourier transform,and the characteristic response matrix was constructed according to the characteristics.The image reconstruction model was constructed and reconstructed using a linear inverse filter and a support vector regression machine.Using cross-correlation coefficient as evaluation index,the analysis of the reconstruction model shows that the response duration is 0.55 s,the response delay time is 0.01 s,and the frequency band is 20-150 Hz.The characteristic response matrix can better characterize visual image stimulus.(4)Tested the image reconstruction model construction algorithm and analyzed the reconstruction results.The software interface was designed,which integrates two subsystems and five modules,and an image reconstruction software system based on LFP signal of pigeon's OT region was realized.The neuron signals of 6 images of 6 pigeons were collected and tested by the image reconstruction model algorithm.The average correlation coefficient of the reconstructed images was 0.9226±0.0212.The results show that the reconstruction model could reconstruct the image stimulus better.The average cross-correlation coefficients of the reconstructed images of 6 pigeons were analyzed,which were 0.9357±0.0224,0.9263±0.0221,0.9281±0.0228,0.9146±0.0165,0.9145±0.0211,0.9165±0.0157,and the mean square error of the average cross-correlation coefficient of the six pigeons was 0.0087.The results show that the pigeon's own difference has no effect on the image reconstruction of the reconstruction model.The average cross-correlation coefficients of the reconstructed images of the two construction algorithms are analyzed,which were 0.9252±0.0227,0.9200±0.0195,respectively.and the mean square error of the average cross-correlation coefficient of the two reconstruction algorithms was 0.0036.The results show that the difference of construction algorithm has no effect on the image reconstruction of the system.