Sleep problems have received much attention from the society and more and more studies have focused on sleep quality and its related diseases, but there is still a lack of high-precision sleep stage identification methods in the field of daily sleep recording. In this paper, we propose an attention mechanism-based sleep stage recognition method, namely the Sleep Stage Attention Model (SSAM), for classifying sleep stages. SSAM uses deep networks for sleep electroencephalographic signal (EEG) features to obtain high-precision sleep signal features. Secondly, we use the attention mechanism to expand the above features for sleep periods to enhance the specific sleep stage. The semantic information of the sleep signal, and the above feature set is transmitted to the subsequent sleep stage recognizer for sleep stage classification. We conducted experiments on datasets Sleep-EDF-20 and Sleep-EDF-78, the experimental results show that our proposed method can obtain higher prediction accuracy than the baseline, which indicates that the use of the attention mechanism effectively improves the sleep stage recognition accuracy.