Pedestrian tracking is an important task in the field of computer vision and is the basis for other advanced vision tasks such as human pose estimation, motion recognition and behavioural analysis, and is widely used in emerging areas such as autonomous driving, intelligent security and service robotics. The detection-based tracking framework in this paper relies heavily on pedestrian detection, and excellent detection algorithms can significantly improve tracking performance. To address the problem of poor pedestrian detection accuracy in the presence of small targets, occlusion and congestion, this paper designs a joint attention module, combines this module with the YOLO-v4 target detection model to design a joint attention-based pedestrian detection algorithm. It conducts experiments on the OTB100- UP&T pedestrian detection dataset. The experimental results demonstrate the effectiveness of the algorithm.