Classical and subsequence dynamic time warping for recognition of rigid body motion trajectories

In many robotic applications, the motions of a human and a robot are recognized by studying of their motion trajectories. Hence, motion trajectory recognition is important in human and robot movement analysis. In this paper, the recognition of six degrees-of-freedom rigid body motion trajectory of an object is studied. The three-dimensional measured position trajectories of LED markers attached to rigid body are transformed to the time-based invariant representation of the rigid body motion trajectories. The main objective of this paper is to evaluate the performance of classical and subsequence dynamic time warping algorithm on the recognition of the rigid body motion trajectories. The experimental results show that the use of the length of shortest warping path in the calculation of DTW distances is not significantly better than the use of only the length of model signal for the nine artificial motions used in experiments. However, the used method promises to improve recognition of more complex everyday motions. Additionally, the results indicate that the classical DTW algorithm gives more meaningful results than subsequence DTW algorithm for the available recorded motions.