Research On Encoding And Decoding Of Human Brain Visual Information Based On Graph Convolutional Network And Recommendation Algorithm

The human brain is a very complex and important organ in the human body,and it is very meaningful to study the function of the human brain.Visual information is the main information source of human cognition,and its operation mechanism has always been a research hotspot in brain science.Functional Magnetic Resonance Imaging(f MRI)is currently the most important research tool for observing human brain activity.Aiming at the decoding of visual information in the human brain,this thesis mainly uses the graph convolutional network model to decode the f MRI signals obtained by the human brain under static image stimulation.Taking the f MRI data of visual information in the human brain as the research object,the f MRI data is preprocessed and feature selection is performed,and then the graph data is constructed in two different ways,and then the Graph Convolutional Network(GCN)is established.Model and Support Vector Machines(SVM)model for category decoding.Aiming at the coding of visual information in the human brain,this thesis mainly studies the relevant knowledge and methods of the recommendation system by comparing the coding process of the stimulus image to the process of the user purchasing goods in the recommendation system.The main work of this thesis includes:1.A method for decoding visual information of human brain based on graph convolutional neural network model is proposed.Firstly,data preprocessing and feature selection are performed on the f MRI data,and then the structure and information of voxels are used to set up nodes and edges in two different ways to establish graph structure data,and establish graph convolutional neural network model and SVM.model to decode.The experimental results show that for the f MRI signals obtained by the human brain under the stimulation of visual images,the models constructed by the two methods have good classification effects.2.A method for encoding visual information of human brain based on recommendation algorithm is proposed.In this thesis,the recommendation algorithm is innovatively connected with the f MRI signal research,the voxels in the f MRI data are analogized to the commodities in the recommendation algorithm,and the stimulation images are regarded as the users,and then the stimulation images can be encoded as voxels in the f MRI information encoding.The process is analogous to the process of users purchasing goods in the recommendation algorithm,and then using the construction method of the network model in the recommendation algorithm,through the recall operation based on the similarity of the items and the sorting regression operation of the sorting layer,the encoded signal of the human brain visual image is obtained.Finally,the encoded analysis of stimulus images is realized.