In this paper, we propose a rule-based system for semantically understanding and analyzing the motion of the trajectories of the human activity. The proposed system can be used as a preprocessing phase for enhancing the object detection process. Detected trajectories are classified into three categories; normal, semi-normal and abnormal trajectories according to the distances between their adjacent points. Abnormal trajectories are removed from the trajectory space. Semi-normal trajectories are broken into small normal trajectories that are linked later to form a longer normal trajectory. The proposed system does not assume a specific trajectory length and hence is more generic than similar trajectory enhancement approaches. The effectiveness of the proposed approach is demonstrated through several experimental results using known human motion datasets.