Research And Implementation Of Image Cognitive FMRI Data Analysis Methods

The image is the carrier of the cognition of external things,improving machine vision performance has long been a popular problem in image processing and pattern recognition fields, the human visual system with excellent cognitive abilities attracted the attention of researchers. We can reverse the cognitive mechanisms of human brain based on fMRI of this non-destructive research techniques.The main content of this paper is to analysis the image cognitive experimental data were in functional magnetic resonance environment, analysised the knowledge of mechanisms of human brain images, used to improve machine vision performance.The main research work are as follows: (1)fMRI data pre-processing, to minimize the individual differences between subjects to eliminate the noise data of the machine, the human heartbeat and respiration, and data standardization. (2)Used generalized linear models, combined with existing bio-priori knowledge, extracted the ROI (Region Of Interest) as a source of brain regions signal. (3)Attempt to make correlation coefficient metric rule on the ROI of BOLD timing signal, establish image cognition functional connectivity matrix, as the high-level features of the image cognition. (4)Extract the image’s color histogram and texture features as low-level features. Took canonical correlation analysis on it with high-level features, preliminary got the human brain’s cognitive model of low-level features.This experiment can be used as an instance of Chinese cognitive experiments of fMRI images,and get the color-related and texture-related functional brain regions. Mainly distributed in Postcentral Gyrus,Cuneus,Lingual Gyrus,Inferior Occipital Gyrus, Inferior Parietal Lobule,Middle Occipital Gyrus,Brodmann area 18&19.And get the interrelatedness of the visual brain regions with low-level feature attributes.