Scaling and time warping in time series querying
VLDB '05 Proceedings of the 31st international conference on Very large data bases
Using object and trajectory analysis to facilitate indexing and retrieval of video
Knowledge-Based Systems
Scaling and time warping in time series querying
The VLDB Journal — The International Journal on Very Large Data Bases
Hi-index | 0.00 |
Given two or more video sequences containing similarhuman activities (running, jumping, etc.) we want to devisea method which extracts spatio-temporal signatures associatedwith these activities, compares these signatures, andaligns key positions of different videos. In this paper weintroduce a method which, in conjunction with a numberof hypotheses, allows the analysis of the motion of specificbody parts and extracts their 2D (image plane) time-varyingtrajectories as well as their 3D trajectories. Two such trajectoriesrecovered from two different videos have differentcharacteristics. We develop a curve registration techniquewhich consists of estimating a transformation mapping onetime-basis (of the first curve) onto another time-basis (thesecond curve). We also analyse in depth the conditions underwhich such curve registration techniques are valid. Finally,we show results with two similar athletic events performedby two different athletes.