A novel human motion recognition method based on eigenspace

  • Authors:

  • Abdunnaser Diaf;Riadh Ksantini;Boubakeur Boufama;Rachid Benlamri

  • Affiliations:

  • University of Windsor, Windsor, ON, Canada;University of Windsor, Windsor, ON, Canada;University of Windsor, Windsor, ON, Canada;Lakehead University, Thunder Bay, ON, Canada

  • Venue:
  • ICIAR'10 Proceedings of the 7th international conference on Image Analysis and Recognition - Volume Part I
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper proposes a novel and robust appearance-based method for human motion recognition based on the eigenspace technique. This method has three main advantages over the existing appearance-based methods. First, the Linear Discriminant Analysis (LDA) is used for dimensionality reduction and eigenspace generation, while preserving maximum separability between classes. Second, by combining a novel centering technique with an incremental procedure, the motion data becomes more concise, expressive, and less confused. Third, data storage is greatly enhanced by using a directed acyclic graph (DAG) structure based on Euclidean distance between projected data. The method is rigorously trained and tested using KTH dataset which contains a large number of motion videos partitioned into six human motions. The experimental results are very promising yielding an average recognition rate of 94.17%.