A Study Of Basic Human Movement Recognition Based On Neuroimaging Data

Human vision is an important way to perceive the external world,and the recognition mechanism of vision has gained great attention from researchers.Now there is a great challenge to utilize computer vision technology to simulate the human brain to analyze and detect the outside world,especially in the study of human behavior due to the problem in efficiency and accuracy.Functional magnetic resonance imaging is a non-invasive technology and it can obtain the functional imaging data of high resolution images.How to identify different human movements by using the functional imaging of human brain was a main issue in our study.To clarify the way how relevant regions in brain cortex process movement information will promote the study on feature extraction and recognition of human behavior in the field of computer vision.It will also provide a theoretical basis for the simulation of human brain by computers,and improve the accuracy of the identification of human activities.The current study explored the roles of regions sensitive to human body form and motion in the decoding of movements and showed the correlation between those regions by time and spatial analysis.Participants were presented four categories of human movements(jump,run,skip and walk)in a blocked fMRI experiment.Support vector machine based on multi-voxel pattern analysis(MVPA)was adopted to assess whether different movements could be discriminated from the activity of four regions.We found that movement-specific information was represented in both human body sensitive areas,extrastriate body area(EBA)and motion sensitive areas,posterior superior temporal sulcus(pSTS)and human middle temporal complex(hMT+).Additionally,a further functional connectivity analysis using EBA as a seed was conducted and it suggested that EBA showed a task-modulated functional connectivity with multiple areas that were involved in the behavior perception and motor control.Human motion processing appeared to be completed in a distributed network.Then the result of dynamic causal model supported our hierarchical processing model for human movements.The occipito-temporal cortex may perform the initial processing of human motion information extracting,and then transformed it to interconnected areas for a further utilization.