AVSP 2003 - International Conference on Audio-Visual Speech Processing September 4-7, 2003 St. Jorioz, France

Identification of Synthetic and Natural Emotional Facial Expressions

Jari Kätsyri, Vasily Klucharev, Michael Frydrych, Mikko Sams

Laboratory of Computational Engineering, Helsinki University of Technology, Finland

Identification of emotional expressions of a Talking Head (TH) were evaluated and compared to that of natural faces. In addition, the effect of static (pictures) and dynamic (video sequences) stimuli was studied. Natural stimuli consisted of six basic emotional expressions (anger, disgust, fear, happiness, sadness and surprise). Two expression sets were selected from both Ekman-Friesen facial affect pictures [1] and Cohn-Kanade database [2]. In addition, two new natural expression sets were recorded in our laboratory. Synthetic expressions were created by our new TH [3], both with and without facial texture. Preliminary results indicate that the TH expressions, except fear, were identified as expected. Overall level of identification of TH stimuli was below those of natural ones. Of all used stimuli, happiness was identified the best and fear the worst. Natural static and dynamic expressions were identified equally well. However, the dynamic expressions of the TH were identified significantly more accurately than the static ones.

References

1. Ekman, P. and Friesen, W., Pictures of Facial Affect, Consulting Psychologists Press, Palo Alto, CA, 1978.
2. Kanade, T., Cohn, J. F., Tian, Y., "Comprehensive Database for Facial Expression Analysis", 4th IEEE Int. Conference on Automatic Face and Gesture Recognition Proc., 2000, p 46 - 53.
3. Frydrych, M., Kätsyri, J., Dobsík, M., Sams, M., ``Toolkit for animation of Finnish Talking Head”, AVSP 2003.

Full Paper

Bibliographic reference.  Kätsyri, Jari / Klucharev, Vasily / Frydrych, Michael / Sams, Mikko (2003): "Identification of synthetic and natural emotional facial expressions", In AVSP 2003, 239-243.