Author(s): Ploetz T, Chen C, Hammerla N, Abowd G
Abstract: The common practice of manual synchronization of body- worn, logging accelerometers and video cameras is imprac- tical for integration into everyday practice for applications such as real-world behavior analysis. We significantly ex- tend an existing technique for automatic cross-modal syn- chronization and evaluate its performance in a realistic ex- perimental setting. Distinctive gestures, captured by a cam- era, are matched with recorded acceleration signal(s) using cross-correlation based time-delay estimation. PCA-based data pre-processing makes the procedure robust against ori- entation mismatches between the marking gesture and the camera plane. We evaluated five different marker gestures and report very promising results for actual use.
|
Dr Thomas Ploetz
|
|