Research On Reconstruction Of Color Character Based On Spike Signal Of Pigeon Optic Tectum Neurons

It is an effective way to conduct neuroscience research by reconstructing sensory information.Color is a natural attribute of objects and environment.Reconstruction of color character stimuli using neuron action potential is of great significance for analyzing the processing mechanism of color information about visual systems.Pigeons have excellent color information perception,and the Optic Tectum(OT)is the key nerve nucleus for pigeons to process color information.Based on the purpose of analyzing the encoding mechanism of neurons by means of reconstruction,different color character stimulation modes were designed,and the action potential response signal(Spike)of pigeon optic tectum neurons were collected,the characteristics of neurons receptive field under different color chessboard stimuli were analyzed.Then the reconstruction model of color information was constructed based on the neuron response,and different color characters were reconstructed.Finally,the neural mechanism of pigeon OT neurons for color information processing was analyzed.The main work and research results of this paper are as follows:(1)Analysis of the color receptive field characteristics of neurons in pigeon OT region.Receiving stimuli about color information in the receptive field of neurons in the OT region of pigeon is a prerequisite for the effective release of neuroelectric signals.In this paper,red(R),green(G),blue(B)in HSV color space model were selected as basic colors.Then,red,green and blue checkerboard stimulus under a white background,green and blue checkerboard stimulus under a red background,red and blue checkerboard stimulus under a green background,red and green checkerboard stimulus under a blue background were designed.And the receptive field were determined by a spike-triggered average(STA)combined with a Gaussian fitting.Based on the comparison of the receptive fields stimulated by red chessboard under white background,it is found that the relative deviation of the central position of the receptive fields stimulated by green chessboard under white background is 2.1902±0.3656 mm,the relative deviation of the radius is 3.3648±0.1982 mm,and the relative deviation of the central position of the receptive fields stimulated by blue chessboard under white background is 2.0690±0.3953 mm,and the relative deviation of the radius is 2.9564 ± 0.6217 mm.For checkerboard experiments under different backgrounds,it is found that the minimum relative deviation of the central position of the receptive field is 1.4560±0.2317 mm,the maximum deviation of the central position is 2.9085 ± 0.4402 mm,the minimum deviation of the radius is 1.2875 ± 0.4525 mm,and the maximum deviation of the radius is 4.3955 ± 0.5562 mm.The above analysis shows that the position and size of the receptive field have changed under different color stimuli,which requires adjusting the position of the stimulus in time to overcome the influence of the position and size of the receptive field on the experimental data.(2)The response characteristics of neuron action potential were extracted and the reconstruction model was constructed.Extraction of neuron response features that can effectively represent color visual stimuli is the prerequisite of reconstruction.The stimuli mode of red characters under white background(W-R),green characters under white background(W-G),blue characters under white background(W-B),green characters under red background(R-G),blue characters under red background(R-B),red characters under green background(G-R),blue characters under green background(G-B),red characters under blue background(B-R)and green characters under blue background(B-G)for B,N,hollow triangle,hollow circle and cross were designed.The characteristics of firing rate of neuron cluster were extracted and the dimension of response characteristic matrix was reduced by PCA.Then the response feature matrix was combined with Linear Inverse Filter algorithm(LIF),random forest algorithm(RF)and Convolutional Neural Networks(CNN)to construct reconstruction model for color characters.(3)The model parameters were selected,the color characters were reconstructed,and the color information processing mechanism of neurons in the OT region was analyzed.The selection of reconstruction model parameters is helpful to improve the performance of the reconstruction model.For the neuron channels involved in reconstruction,the neuron channels were selected by means of information theory and reconstruction accuracy.The stimulus duration(T),start time(delay)and time window(bin)were selected by the joint method of grid search method and cross test method,and the better reconstruction model was constructed by using the selected parameters.For the analysis of reconstruction results about color characters,firstly,the reconstruction models were used to reconstruct the color characters.The result shows that the highest reconstruction accuracy of LIF model is 0.9778,the highest reconstruction accuracy of RF model is 0.9675,and the highest reconstruction accuracy of CNN model is 0.9822.This shows that the feature of firing rate can represent the color information.Then,the analysis by means of ANOVA indicates that there is no significant difference between the reconstruction results about color characters under the white background.And the results of character reconstruction under different backgrounds were analyzed.The results indicates that when the characters are red,there is no significant difference for the results of reconstruction under different color background.When the character is blue,the red background weakens the response of blue stimuli.When the character is green,the blue background enhances the visual effect of green.Finally,the analysis of the reconstruction results for character stimuli under different color combinations shows that there are significant differences between reconstruction results of blue characters under red background(R-B),red characters under blue background(B-R)and reconstruction results of blue characters under green background(G-B),blue characters under green background(G-B),green characters under red background(R-G).It indicates Blue and red affect each other and weaken the intensity of stimuli.