Why are some emotions more difficult to recognise than others?
We think that part of this difficulty is due to how consistent the emotions are expressed across different people. That is, the more similar (i.e., consistent) people produce a particular expression, the less variability there are in that expression, and thus, the easier the expression is recognised.
For example, we think that ‘happy’ is an emotion that is quite similarly expressed across individuals (e.g., most individuals expressing happiness will have squinty eyes, wrinkled nose, and a wide grin or smile). Conversely, we think ‘sarcastic’ might be an example of an emotion that would have high variability in how it is expressed across individuals (e.g, some might roll their eyes, some might widen their eyes, while some might have a deadpan expression). For that reason, we think that ‘happy’ would be easier to recognise than ‘sarcastic’.
What did we find?
We compared three computational measures to determine which would best fit human recognition data, and we found that one measure, which we call ‘CoM-CoM’, provides the best fit. Essentially, this measure is computed by reducing the audio-visual clip of the expression into a matrix to its “center”. We do this for all the clips of the same emotion, and obtain the center of the centers.
Publications
Paper in preparation.