The non-stationary of the motor imagery electroencephalography(MI-EEG) signal is one of the main limitations for the development of motor im-agery brain-computer interfaces(MI-BCI).The non-stationary of the MI-EEG signal and the changes of the experimental environment make the feature dis-tribution of the testing set and training set deviates,which reduces the classification accuracy of MI-BCI.In this paper,we propose a Kullback-Leibler diver-gence (KL)-based transfer learning algorithm to solve the problem of feature transfer,the proposed algorithm uses KL to measure the similarity between the train-ing set and the testing set,adds support vector ma-chine (SVM) classification probability to classify and weight the covariance,and discards the poorly per-forming samples.The results show that the proposed algorithm can significantly improve the classification accuracy of the testing set compared with the tradi-tional algorithms,especially for subjects with medium classification accuracy.Moreover,the algorithm based on transfer learning has the potential to improve the consistency of feature distribution that the traditional algorithms do not have,which is significant for the ap-plication of MI-BCI.