Motion capture data that describe human movements are gaining popularity in a number of application domains, e.g., sports, medicine, or security. Consequently, a lot of current research efforts is dedicated to efficient and effective processing of the motion capture data. The DISA research group at FI MU has recently proposed a new compact representation of motion data, denoted as Motion Words. It has been demonstrated that the Motion Words can significantly speed up motion processing, but the semantic expressiveness of this representation has not been fully explored yet. In particular, the utilization of Motion Words may be problematic when we need to compare motions performed at different speed.

The objective of this thesis is to examine the behavior of the Motion Words on motions that are performed with different speed. The student will analyze the behavior of Motion Words both on real-world motion datasets and on artificial datasets of fast and slow motions. Next, the student will propose an extension of the Motion-Word-based similarity searching that will allow to find semantically related motions regardless of their speed.