 |
Effect of Initial HMM Choices in Multiple Sequence Training for Gesture Recognition
Liu, Nianjun, Davis, Richard I. A., Lovell, Brian C. and Kootsookos, Peter J. (2004). Effect of Initial HMM Choices in Multiple Sequence Training for Gesture Recognition. In: P. Srimani, Proceedings of the International Conference on Information Technology: Coding and Computing. International Conference on Information Technology, Las Vegas, (608-613). 5-7 April, 2004.
|
|
| |
| Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials) |
| Name |
Description |
MIMEType |
Size |
Downloads |
itccDavisLiuLove.pdf
|
itccDavisLiuLove.pdf |
application/pdf |
186.84KB |
622 |
| Author(s) |
Liu, Nianjun
Davis, Richard I. A.
Lovell, Brian C.
Kootsookos, Peter J.
|
| Title of paper |
Effect of Initial HMM Choices in Multiple Sequence Training for Gesture Recognition
|
| Conference name |
International Conference on Information Technology
|
| Conference location |
Las Vegas
|
| Conference dates |
5-7 April, 2004
|
| Proceedings title |
Proceedings of the International Conference on Information Technology: Coding and Computing
|
| Editor(s) |
P. Srimani
|
| Place published |
Los Alamitos, California
|
| Publisher |
The Institute of Electrical and Electronics Engineers Computer Society
|
| Publication date |
2004
|
| Volume number |
1
|
| ISBN |
0-7695-2108-8
|
| Start page |
608
|
| End page |
613
|
| Total pages |
6
|
| Language |
eng
|
| Abstract/Summary |
We present several ways to initialize and train Hidden Markov Models (HMMs) for gesture recognition. These include using a single initial model for training (reestimation), multiple random initial models, and initial models directly computed from physical considerations. Each of the initial models is trained on multiple observation sequences using both Baum-Welch and the Viterbi Path Counting algorithm on three different model structures: Fully Connected (or ergodic), Left-Right, and Left-Right Banded. After performing many recognition trials on our video database of 780 letter gestures, results show that a) the simpler the structure is, the less the effect of the initial model, b) the direct computation method for designing the initial model is effective and provides insight into HMM learning, and c) Viterbi Path Counting performs best overall and depends much less on the initial model than does Baum-Welch training.
|
| Subjects |
280207 Pattern Recognition
E1
|
| Keyword(s) |
iris-research
gesture recognition
hidden Markov models
|
|
|
|
