Automatic imitation assessment in interaction

  • Authors:
  • Stéphane Michelet;Koby Karp;Emilie Delaherche;Catherine Achard;Mohamed Chetouani

  • Affiliations:
  • Institute of Intelligent Systems and Robotics, University Pierre and Marie Curie, Paris, France;Institute of Intelligent Systems and Robotics, University Pierre and Marie Curie, Paris, France;Institute of Intelligent Systems and Robotics, University Pierre and Marie Curie, Paris, France;Institute of Intelligent Systems and Robotics, University Pierre and Marie Curie, Paris, France;Institute of Intelligent Systems and Robotics, University Pierre and Marie Curie, Paris, France

  • Venue:
  • HBU'12 Proceedings of the Third international conference on Human Behavior Understanding
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Detecting social events such as imitation is identified as key step for the development of socially aware robots. In this paper, we present an unsupervised approach to measure immediate synchronous and asynchronous imitations between two partners. The proposed model is based on two steps: detection of interest points in images and evaluation of similarity between actions. Firstly, spatio-temporal points are detected for an accurate selection of the important information contained in videos. Then bag-of-words models are constructed, describing the visual content of videos. Finally similarity between bag-of-words models is measured with dynamic-time-warping, giving an accurate measure of imitation between partners. Experimental results obtained show that the model is able to discriminate between imitation and non-imitation phases of interactions.