“…Notably, due to its simplicity and effectiveness, the shape of a trajectory is one of the commonly used descriptions for trajectory analysis and anomaly detection [7][8][9], and our paper focuses on shape analysis and anomaly detection. The shape of a motion trajectory is classically characterized by a series of tangential angles at the positions of the corresponding moving object (for the sake of easy description, in this paper, we call these positions as the sample points of the trajectory), and these angles are usually modeled by using the von Mises distribution to represent the trajectory [7,8,10]. However, despite the fact that the angle attribute is largely helpful for a lot of practical applications, the use of only this single attribute cannot discriminate, for instance, whether a person is walking or running along a same path in an automated surveillance system.…”