Protractor: a fast and accurate gesture recognizer

  • Authors:
  • Yang Li

  • Affiliations:
  • Google Research, Mountain View, CA, USA

  • Venue:
  • Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
  • Year:
  • 2010

Quantified Score

Hi-index 0.01

Visualization

Abstract

Protractor is a novel gesture recognizer that can be easily implemented and quickly customized for different users. Protractor uses a nearest neighbor approach, which recognizes an unknown gesture based on its similarity to each of the known gestures, e.g., training samples or examples given by a user. In particular, it employs a novel method to measure the similarity between gestures, by calculating a minimum angular distance between them with a closed-form solution. As a result, Protractor is more accurate, naturally covers more gesture variation, runs significantly faster and uses much less memory than its peers. This makes Protractor suitable for mobile computing, which is limited in processing power and memory. An evaluation on both a previously published gesture data set and a newly collected gesture data set indicates that Protractor outperforms its peers in many aspects.