 Multisource transfer learning, MSTL, has been widely used in motor imagery, MI, based brain-computer interfaces, BCIs, as it can reduce individual differences between subjects. However, existing MSTL methods often fail to capture the differences between subjects, resulting in poor performance. In this paper, we propose two new MSTL algorithms, transfer joint matching, TJM, and weighted TJM, WTJM, to better align the data distributions between subjects. These algorithms are designed to select the best samples from each subject and then use them to generate a shared representation. This shared representation is then used to classify the data from both subjects. Our experiments show that the proposed algorithms significantly outperform existing MSTL methods in terms of accuracy. This article was authored by Fulinwe, Shuayunsu, Tianyuanjia, and others. We are article.tv, links in the description below.