In view of this problem of low accuracy and poor universality due to lack of other relevant information, a motor imagination classification method based on spatial and frequency domain features was proposed. According to event related synchronization, event related desynchronization, and the contralateral mapping mechanism during moving, the combined features of right and left hands extracted by common spatial pattern and power spectrum density from the channels of C3 and C4 were classified using support vector machine. The results show that the combining features solved the problem of common spatial pattern is sensitive to noise, and made up for the lack of frequency features information. A higher classification result and generalization was also achieved, the classification accuracy was up to 91.3%, which verified the effectiveness of the method proposed in this paper.