A Comparative Evaluation of Some Gesture Recognition Algorithms
(2011)
Files
Beuvens-Interact2011.docx
Open Access - Microsoft Word XML
- 1.08 MB
Details
- Authors
- Abstract
- Several algorithms have been developed for pen-based gesture recognition. Yet, selecting which one is the most appropriate in which context is limited by several shortcomings: they are hard to compare to each other since they operate with dif-ferent parameters, determining which one is the most suitable in which situation is a research problem, their performance largely vary depending on contextual pa-rameters that are hard to predict, their fine-tuning in a real interactive application remains a challenge. In order to address these shortcomings, this paper reports on an experimental study that involved 30 participants testing 4 algorithms for pen-based gesture recognition based on a data set of 26 letters, 10 figures, 17 action commands, and 8 geometrical shapes. The results of this study suggests that the four algorithms can be effectively compared with respect to the following pa-rameters that provide some guidance for their selection: resampling, rescaling, ro-tation, kNN multiplicity, multi-stroke, Multi-stroke combined with resampling, pressure, 3 different normalizations according to 3 different metrics, and adapted costs matrix
- Affiliations
Citations
Beuvens, F. (2011). A Comparative Evaluation of Some Gesture Recognition Algorithms. https://hdl.handle.net/2078.5/274735