A Segmentation Algorithm For the Comparison of Human Limb Trajectories
Marc Pomplun, Maja J. Matarić · 2001
In this article, we present an analytical approach for movement evaluation and comparison. We develop a joint-space-based segmentation and comparison algorithm that calculates pairwise similarity of arm movement trajectories. To validate the algorithm, we collected sixty sample arm movement sequences from a subject imitating video demonstrations. We present the performance of our algorithm on that data set, and discuss the contribution of each of its components. SEGMENTATION ALGORITHM 2 A Segmentation Algorithm for the Comparison of Human Limb Trajectories Comparing and evaluating movement is a known diÆcult problem in psychophysics, as well as in robotics. Our particular area of interest is in motor skill learning by imitation, both in humans and humanoid robots (Byrne & Russon, 1998; Mataric, 2001), where the need for analytical movement comparison is particularly acute. Our earlier work (Mataric & Pomplun, 1997, 1998) in the area of psychophysics of imitation involved studying the attentional behavior of people watching movements for subsequent imitation. In more recent experiments, we aimed to answer more precise questions about the quality of the subjects ' imitation under various conditions, by recording the demonstrator's and the