Automatic Facial Expression Analysis

Abstract

Automatic Facial Expression Recognition systems have come a long way since the earliest approaches in the early 1970s. We are now at a point where the earliest systems are commercially applied, most notably the smile detectors in digital cameras. But although facial expression recognition is maturing as a research field, it is far from finished. New techniques continue to be developed on all aspects of the processing pipeline: From face detection, via feature extraction to machine learning. Nor is the field blind to the progress made in the social sciences with respect to emotion theory. Gone are the days that people only tried to detect six discrete expressions that were turned-on or off like the switching of lights. The theory of Social Signal Processing now complements classical emotion theory, and modern approaches dissect an expression into its temporal phases, analyse intensity, symmetry, micro-expressions and dynamic differences between morphologically similar expressions. Brave new worlds are opened up—Automatic Facial Expression Analysis is poised to revolutionalise medicine with the advent of behaviomedics, gaming with enriched player–non-player interactions, teleconference meetings with automatic trust and engagement analysis, and human–robot interaction with robots displaying actual empathy.