1. Sentence:
“In this work we demonstrate that identifying locally similar regions in human motion data can be practical even for huge databases, if medium-dimensional (15–90 dimensional) feature sets are used for kd-tree-based nearest-neighbor searches.” (abstract)
2. Problem:
Motion capture creates large amounts of data and searching through the data to find similar motions in a database can take large amounts of time.
3. Key Idea:
The kd-tree based nearest neighbor algorithm can be used well with searching a database for matching motion data.
4.a. What the paper does:
The paper provides a way using limited number of data points and a kd-tree based nearest neighbor algorithm to find matching motion data in shorter amounts of time than previous methods.
4.b. What it could be used for:
Visualizing many objects at same time that have approximately same motion data or motion reconstruction using a limited number of markers.
5. Resources:
Supplemental Material , Video