Reliable, robust, accurate, and real-time human tracking is essential for mobile robots and intelligent vehicles in real-life applications. 2D LiDAR is considered as the standard sensor for mobile robot navigation as well as human detection and tracking due to its low-cost and usability. However, 2D range limitation and occlusion caused by obstacles, especially dynamic human environments, make it less reliable, robust and accurate for human tracking. This letter introduces a new method for increasing the quality of 2D LiDAR human tracking in cluttered and crowded environments. We combined human content presented by Hall's Proxemics model with the global nearest neighbor to improve accuracy of scan-to-track data association of leg detection. Social dynamic confidence (SDC) factor is generated based on features of human social norms, dynamic metrics and consistency developed in the detection stage. As a result, our proposed method improved multi-object tracking accuracy and runtime 24% and 45%, respectively, against the state-of-the-art joint-leg-tracker technique in crowded and cluttered environments.