Cyclic and Non-Cyclic Gesture Spotting and Classification in Real-Time Applications
Abstract
This paper presents a gesture recognition method for detecting and classifying both cyclic and non-cyclic human motion patterns in real-time applications. The semantic segmentation of a constantly captured human motion data stream is a key research topic, especially if both cyclic and non-cyclic gestures are considered during the humancomputer interaction. The system measures the temporal coherence of the movements being captured according to its knowledge database, and once it has a sufficient level of certainty on its observation semantics the motion pattern is labeled automatically. In this way, our recognition method is also capable of handling time-varying dynamic gestures. The effectiveness of the proposed method is demonstrated via recognition experiments with a triple-axis accelerometer and a 3D tracker used by various performers.
BIB_text
@Article {
author = {Luis Unzueta and Jon Goenetxea},
title = {Cyclic and Non-Cyclic Gesture Spotting and Classification in Real-Time Applications},
pages = {172-181},
volume = {6169},
keywds = {
Human-Computer Interaction, Gesture Spotting, Gesture Recognition, Motion Pattern, Motion Capture
}
abstract = {
This paper presents a gesture recognition method for detecting and classifying both cyclic and non-cyclic human motion patterns in real-time applications. The semantic segmentation of a constantly captured human motion data stream is a key research topic, especially if both cyclic and non-cyclic gestures are considered during the humancomputer interaction. The system measures the temporal coherence of the movements being captured according to its knowledge database, and once it has a sufficient level of certainty on its observation semantics the motion pattern is labeled automatically. In this way, our recognition method is also capable of handling time-varying dynamic gestures. The effectiveness of the proposed method is demonstrated via recognition experiments with a triple-axis accelerometer and a 3D tracker used by various performers.
}
isbn = {978-3-642-14060-0},
date = {2010-07-09},
year = {2010},
}
author = {Luis Unzueta and Jon Goenetxea},
title = {Cyclic and Non-Cyclic Gesture Spotting and Classification in Real-Time Applications},
pages = {172-181},
volume = {6169},
keywds = {
Human-Computer Interaction, Gesture Spotting, Gesture Recognition, Motion Pattern, Motion Capture
}
abstract = {
This paper presents a gesture recognition method for detecting and classifying both cyclic and non-cyclic human motion patterns in real-time applications. The semantic segmentation of a constantly captured human motion data stream is a key research topic, especially if both cyclic and non-cyclic gestures are considered during the humancomputer interaction. The system measures the temporal coherence of the movements being captured according to its knowledge database, and once it has a sufficient level of certainty on its observation semantics the motion pattern is labeled automatically. In this way, our recognition method is also capable of handling time-varying dynamic gestures. The effectiveness of the proposed method is demonstrated via recognition experiments with a triple-axis accelerometer and a 3D tracker used by various performers.
}
isbn = {978-3-642-14060-0},
date = {2010-07-09},
year = {2010},
}
