Sign recognition using constrained optimization

  • Authors:
  • Kikuo Fujimura;Lijie Xu

  • Affiliations:
  • Honda Research Institute;Ohio State University

  • Venue:
  • ACCV'07 Proceedings of the 8th Asian conference on Computer vision - Volume Part II
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Sign recognition has been one of the challenging problems in computer vision for years. For many sign languages, signs formed by two overlapping hands are a part of the vocabulary. In this work, an algorithm for recognizing such signs with overlapping hands is presented. Two formulations are proposed for the problem. For both approaches, the input blob is converted to a graph representing the finger and palm structure which is essential for sign understanding. The first approach uses a graph subdivision as the basic framework, while the second one casts the problem to a label assignment problem and integer programming is applied for finding an optimal solution. Experimental results are shown to illustrate the feasibility of our approaches.