The Facial Action Coding System is built to code the intensity of each action unit. But in studies on emotional facial expressions, how the intensity of global expressions is coded ? I have not found articles that indicate if some rules exist when someone want to measure the global intensity of an expression (e.g. not AU 6+12 separately, but the smile as a whole).
Many thanks for considering my request.
Since you need a coder anyway he can also estimate the intensity using some rating scale. Look for Kring's Facial Expression Coding System (FACES) as an example.
Jeffrey gave an interesting hint at how we can combine individual AU's intensities into a general intensity index. But as he noted, it has limitations. I would add that since FACS intensity levels are not on an interval scale (for example, the C level covers a wider range of movement than the B or the A levels), using a 1 to 5 linear scale to compute intensity values does not represent what was intended in FACS (let alone the fact that the adjectives "slight", "marked", or "severe" can be interpreted differently by different people). Based on the paper by Sayette et al. (2001) in JONB, we used three levels to code AU intensity then we compute an expressivity index for each AU based on intensity and duration of each AU. Once you do this for several AUs you can decide of a way to combine these values. As far as I know there is no gold standard
- Badges
- Science topic