Abstract: Due to the small number of training samples of motor imagery EEG for a single subject and the great individual differences of EEG among different subjects, the accuracy of brain motor imagery task classification is low and the training calibration time is too long. In this paper, a transfer learning method of Euclidean aligned-regular common space pattern is proposed to realize the transfer of samples and features at the same time. Firstly, the data alignment transfer method is used to make the sample distribution more similar among different subjects, and then the regularized transfer method is applied to improve the common spatial pattern and extract more robust spatial feature information. Simulation test on BCI competition motor imagery EEG dataset shows the average classification accuracy of motor imagery tasks for right hand and right foot can reach 87.10%. On the small training sample set, the average classification accuracy of the transfer learning method is 14.43% higher than that of the common space pattern method. For the motor imagery EEG across subjects, the average classification accuracy is improved by 23.5%. The experimental results show that this method can effectively improve the classification accuracy of motor imagery EEG for small sample and cross subject, reduce the training time of model calibration, and improve the generalization ability.