Efficient Distance-Based Gestural Pattern Mining in Spatiotemporal 3D Motion Capture Databases

One of the most fundamental challenges when mining gestural patterns in 3D motion capture databases is the definition of spatiotemporal similarity between two gestural patterns. While time-elastic similarity models such as the Gesture Matching Distance on gesture signatures are able to leverage the spatial and temporal characteristics of gestural patterns, the applicability of such distance-based models in order to analyze large 3D motion capture databases is limited due to their high computational complexity. To this end, we propose a lower bound approximation of the Gesture Matching Distance that preserves the spatiotemporal characteristics and can be utilized in an optimal multi-step k-nearest-neighbor search architecture in order to analyze and mine spatiotemporal databases efficiently. We empirically investigate the performance in terms of accuracy and efficiency based on 3D motion capture databases and show that our lower bound approximation is able to achieve an increase in efficiency of more than one order of magnitude with a negligible loss in accuracy. Our proposal is fundamental for efficient distance-based gestural pattern mining.