Recognizing human motion trajectories in videos plays a crucial role in action recognition. In sign language, these trajectories are primarily exhibited through hand and facial movements across consecutive frames. However, existing continuous sign language recognition technologies typically process each frame independently, leading to ineffective capture of cross-frame motion trajectories and thus affecting gesture recognition outcomes. To address this issue, we propose a Dynamic Trajectory Correlation Module that dynamically computes the correlations between the current frame and the next, to identify the motion trajectories within that spatial area. Experimental results demonstrate that our continuous sign language recognition model achieves state-of-the-art performance on sign language recognition tasks across multiple datasets, including Phoenix-2014, Phoenix-2014T, and CSL-Daily.