To solve the data classification of the functional magnetic resonance imaging(fMRI) signals in the brain-compute interface, the classification method of support vector machine(SVM) using posterior parietal cortex(PPC) as feature selection was presented. First, the data were acquired by the nuclear magnetic device. Next,the data were preprocessed, the voxels of PPC were selected as features, then the peak values and cumulative values of BOLD(blood oxygen level dependent) were selected as the feature extraction. Finally, SVM was used to classify data. The experiment has shown it is viable to select PPC as feature and the classification accuracy using peak value is higher than the classification accuracy using cumulative value.