Moments and wavelets for classification of human gestures represented by spatio-temporal templates
AI'04 Proceedings of the 17th Australian joint conference on Advances in Artificial Intelligence
| Hi-index | 0.00 |
This paper evaluates the efficacy of directionalinformation of wavelet multi-resolution decomposition forhistogram-based classification of human gesturesrepresented by spatio-temporal templates. This templatecollapses temporal component into gesture representationin a way that no explicit sequence matching or temporalanalysis is needed, and characterizes the motion from avery high dimensional space to a low dimensional space.These templates are modified to be invariant totranslation, rotation and scale. Two dimensional, 3 leveldyadic wavelet transform applied on these templatesresults in one low pass subimge and nine highpassdirectional subimages. Histograms of wavelet coefficientsat different scales are compared to establish significanceof available information for classification. Thepreliminary experiments show that while the statisticalproperties of the template provide high level ofclassification accuracy, the available information in highpass or low pass decompositions by itself is not sufficientto provide significant efficiency of accuracy.