 This study examined the effects of intra-artifacts on deep learning-based sleep staging systems. The authors found that intra-artifacts can have a negative impact on the accuracy of the systems and proposed a new method called SOBWT which combines wavelet transform and second-order blind source separation to reduce the negative impact of intra-artifacts. This method was tested on three different datasets and showed improved accuracy when compared to other artifact reduction techniques. This article was authored by Hanyu Zhu, Yonglin Wu, Mingxian, and others.