1F: one accessory feature design for gesture recognizers

One Feature (1F) is a simple and intuitive pruning strategy that reduces considerably the amount of computations required by Nearest-Neighbor gesture classifiers while still preserving the high recognition rate. Performance results are reported for 1F by analyzing a large set of candidate features showing recognition rates of 99% with a peak reduction in computations of 70%. 1F is easy to implement, flexible with respect to the choice of the feature, and exploits the intuition of the designer by exposing clear innerworkings. Author Keywords Gesture recognition; nearest neighbor; feature selection; pruning; training set; classification; comparing classifiers; gesture descriptors ACM Classification Keywords H.5.2 Information Interfaces and Presentation (e.g. HCI): User Interfaces - Input devices and strategies; I.5.2 Pattern recognition: Design methodology - Classifier design and evaluation, Feature evaluation and selection General Terms Algorithms, Design, Experimentation.

• Type: Journal
• Year: 2012
• Where: IUI
• Authors: Radu-Daniel Vatavu
• Keywords: Design Methodology, Feature Evaluation, Gesture Recognition, IUI 2012, Software Engineering