Classification Research Of Brain Cognitive Status Based On MPCA

The cognitive neuroscience research is to reveal the brain cognitive principle. Thekey point of classification research of the brain cognitive state is to construct a goodpattern recognition system, in order to distinguish different cognitive activities.Currently brain fMRI data dimensions and quantities become larger. But because of lotsof calculation cost and short of enough effective features, the traditional PCA algorithmcan’t be applied widely.Based on the traditional PCA algorithm based on the vector and the currentpopular tensor analysis, this paper proposes a new principal component analysis methodbased on the tensor model--multilinear principal component analysis (MPCA). Thismethod will extract the features of tensor objects used as the substitute of original highdimensionality data, and gain the good effect of features extracting. This paper usesMPCA algorithm to extract features, and feature sub-space from LDA is calculated onthe basis of MPCA. Meanwhile, the two feature sub-space from MPCA and LDA arefused, and the fusion feature space is acquired. After training samples and test samplesare respectively projected towards the fusion feature space, recognition feature areaccordingly gained. At last nearest neighbor rule is utilized in the brain cognitive statusclassification. Experiments show that this algorithm has good recognition effect.