by RB Rusu, J Bandouch, F Meier, I Essa and M Beetz
Abstract:
This article investigates the recognition of human actions from 3D point clouds that encode the motions of people acting in sensor-distributed indoor environments. Data streams are time-sequences of silhouettes extracted from cameras in the environment. From the 2D silhouette contours we generate space-time streams by continuously aligning and stacking the contours along the time axis as third spatial dimension. The space-time stream of an observation sequence is segmented into parts corresponding to subactions using a pattern matching technique based on suffix trees and interval scheduling. Then, the segmented space-time shapes are processed by treating the shapes as 3D point clouds and estimating global point feature histograms for them. The resultant models are clustered using statistical analysis, and our experimental results indicate that the presented methods robustly derive different action classes. This holds despite large intra-class variance in the recorded datasets due to performances from different persons at different time intervals.
Reference:
Human Action Recognition using Global Point Feature Histograms and Action Shapes (RB Rusu, J Bandouch, F Meier, I Essa and M Beetz), In Advanced Robotics journal, Robotics Society of Japan (RSJ), 2009.
Bibtex Entry:
@article{rusu_human_2009,
author = {RB Rusu and J Bandouch and F Meier and I Essa and M Beetz},
title = {Human Action Recognition using Global Point Feature Histograms and
Action Shapes},
journal = {Advanced Robotics journal, Robotics Society of Japan ({RSJ)}},
year = {2009},
abstract = {This article investigates the recognition of human actions from {3D}
point clouds that encode the motions of people acting in sensor-distributed
indoor environments. Data streams are time-sequences of silhouettes
extracted from cameras in the environment. From the {2D} silhouette
contours we generate space-time streams by continuously aligning
and stacking the contours along the time axis as third spatial dimension.
The space-time stream of an observation sequence is segmented into
parts corresponding to subactions using a pattern matching technique
based on suffix trees and interval scheduling. Then, the segmented
space-time shapes are processed by treating the shapes as {3D} point
clouds and estimating global point feature histograms for them. The
resultant models are clustered using statistical analysis, and our
experimental results indicate that the presented methods robustly
derive different action classes. This holds despite large intra-class
variance in the recorded datasets due to performances from different
persons at different time intervals.},
}