Recognizing 3D human motions using fuzzy quantile inference

  • Authors:

  • Mehdi Khoury;Honghai Liu

  • Affiliations:

  • Institute of Intelligent Systems and Robotics, Department of Creative Technologies, University of Portsmouth, Portsmouth, United Kingdom;Institute of Intelligent Systems and Robotics, Department of Creative Technologies, University of Portsmouth, Portsmouth, United Kingdom

  • Venue:
  • ICIRA'10 Proceedings of the Third international conference on Intelligent robotics and applications - Volume Part I
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

Fuzzy Quantile Inference (FQI) is a novel method that builds a simple and efficient connective between probabilistic and fuzzy paradigms and allows the classification of noisy, imprecise and complex motions while using learning samples of suboptimal size. A comparative study focusing on the recognition of multiple stances from 3d motion capture data is conducted. Results show that, when put to the test with a dataset presenting challenges such as real biologically noisy" data, cross-gait differentials from one individual to another, and relatively high dimensionality (the skeletal representation has 57 degrees of freedom), FQI outperforms sixteen other known time-invariant classifiers.