• Increase font size
  • Default font size
  • Decrease font size

Facial Expression Recognition

E-mail Print
The component recognizes facial expression in real-time by localizing and tracking facial movements, based on the appearance of the expression of a person when interacting with a camera, and providing feedback about recognised emotions based on dimensional or Ekmanian emotions. Facial features movements and facial animation parameters, relevant to facial affctive expressions detection, are calculated using both emotional and dimensional models. The component can provide a quadrant of a dimensional model of emotions suitable to detect also emotions other than Ekmanians, provided that the rules of a specific emotional model is given to the component.

Besides functional tests, extensive experimentation of Facial expression Recognition in CALLAS is done in edutainment application for Proof-of-Concepts:
  • the Musickiosk : the component was used used for Ekmanian values,enumerated and combined with other input modalities into P-A-D values for emotion estimation. This resulted in guiding the music according to facial expressions: as facial expressions are more negative, music gets more passive, and viceversa if more positive.
  • the Interactive Opera : the component was used for interfering with the story’s plot. Depending on facial expressions of multiple people, each one standing for a different character, the plot is accordingly designed.
Additional reference to the component usage is made in the following CALLAS papers and articles:
  • A non-intrusive method for user focus of attention estimation in front of a computer monitor: Abstract
Reference information about the availability of the component can be obtained by Institute of Communication and Computer Systems - National Technical University of Athens.
Last Updated on Saturday, 15 May 2010 10:24