August 2014
Volume 14, Issue 10
Vision Sciences Society Annual Meeting Abstract  |   August 2014
Recognition of complex and realistic facial expressions of emotion
Author Affiliations
  • Shichuan Du
    Electrical and Computer Engineering, The Ohio State University
  • Pamela Pallett
    Electrical and Computer Engineering, The Ohio State University
  • Aleix M. Martinez
    Electrical and Computer Engineering, The Ohio State University
Journal of Vision August 2014, Vol.14, 1386. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Shichuan Du, Pamela Pallett, Aleix M. Martinez; Recognition of complex and realistic facial expressions of emotion. Journal of Vision 2014;14(10):1386.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Everyday experience shows that naturalistic human facial expressions include complex combinations of signals, i.e., happily surprised or hate (which is a combination of anger and disgust). However, research in facial expressions of emotion has mostly been limited to the study of the prototypical expressions of happiness, surprise, anger, sadness, fear and disgust. This research has shown that human subjects can readily classify these expressions significantly above chance and that this recognition does not decrease even when the images are degraded (e.g., when the resolution of the image is systematically reduced). Here, we ask if this result holds for more realistic, everyday expression, e.g., when observing a facial expression of hate. Participants (n=14) performed an 8-alternative-forced-choice (8AFC) task while observing images of facial expressions of happily surprised, fearfully surprised, sadly angry, happily disgusted, angrily disgusted, awe, hate and neutral. Faces were shown in random order, without repetition, at five different image resolutions; from a high of 240x160 to a low of 15x10 pixels (see supplementary file). To prevent recognition based on previous exposure, an image shown at one resolution was not repeated at a different resolution. Exposure time was 500 ms, followed by a 750ms mask and the 8AFC semantic categorization task. Subjects performed significantly above chance for all categories. Most importantly, the categorization accuracy did not decrease until the images resolution was below 30x20 pixels. This result is consistent with previous experiments on the six prototypical facial expressions described earlier, where it was also observed that recognition stays relatively unchanged even for stimuli of only 30x20 pixels. These results hence suggest that our visual system can readily categorize facial expressions of many complex and naturalistic expressions, way beyond the classical six discussed in the literature.

Meeting abstract presented at VSS 2014


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.