 This paper proposes a convolutional neural network, CNN, based on inter-task transfer learning to improve the accuracy of decoding kinematic information from brain signals. The CNN uses a channelwise variational autoencoder, CVAE, to combine EEG signals from both motor execution, ME, and motor imagery, MI, tasks. This approach allows the model to use a small amount of MI data to improve the classification accuracy of ME data. The proposed method was tested on two datasets and achieved higher classification accuracies than other methods. These results demonstrate the potential of the proposed method for improving the accuracy of BCI systems. This article was authored by Dune Lee, Jihoon Jong, Byeonghu Lee, and others.