Two-stage Recognition of Raw Acceleration Signals for 3-D Gesture-Understanding Cell Phones

Sung-Jung Cho 1 Eunseok Choi 1 Won-Chul Bang 1 Jing Yang 1 Junil Sohn 1 Dong Yoon Kim 1 Young-Bum Lee 1 Sangryong Kim 1
1 SAIT - Samsung Advanced Institute of Technology
Abstract : As many functionalities like cameras and MP3 players are converged to cell phones, more intuitive interaction methods are essential beyond tiny keypads. In this paper, we present gesture-based interactions and their two-stage recognition algorithm. Acceleration signals are generated from accelerometer. At the first stage, they are hierarchically modelled and matched as basic component and their relationships by Bayesian networks. At the second stage, they are further classified by SVMs for resolving confusing pairs. Our system showed enough recognition performance for commercialization; with 100 novice users, the average recognition rate was 96.9% on 11 gestures (digits 1-9, O, X). The algorithms have been adopted in the world-first gesture-recognizing Samsung cell phones since 2005.
Keywords : 3-D gesture recognition acceleration Bayesian networks support vector machines sensorbased interaction
Document type :
Conference papers
Complete list of metadatas

Cited literature [15 references]  Display  Hide  Download
https://hal.inria.fr/inria-00103854
Contributor : Anne Jaigu <>
Submitted on : Thursday, October 5, 2006 - 1:30:20 PM
Last modification on : Thursday, October 5, 2006 - 4:18:17 PM
Long-term archiving on : Tuesday, April 6, 2010 - 6:27:25 PM

File

cr1019154605290.pdf

Identifiers

  • HAL Id : inria-00103854, version 1

Collections

IWFHR10

Citation

Sung-Jung Cho, Eunseok Choi, Won-Chul Bang, Jing Yang, Junil Sohn, et al.. Two-stage Recognition of Raw Acceleration Signals for 3-D Gesture-Understanding Cell Phones. Tenth International Workshop on Frontiers in Handwriting Recognition, Université de Rennes 1, Oct 2006, La Baule (France). ⟨inria-00103854⟩

Export

BibTeX TEI DC DCterms EndNote

Share

Metrics

Record views

623

Files downloads

1038

Contact


archive-ouverte@inria.fr