We have been using similar software (http://www.noldus.com/human-behavior-research/products/facereader http://www.noldus.com/human-behavior-research/products/fa

We have been using similar software (http://www.noldus.com/human-behavior-research/products/facereader http://www.noldus.com/human-behavior-research/products/facer... ) in our research into human observation in context aware computing. On the sensing side it would be interesting to see how this compares. In our experience, the observation goes quite well as long as you have enough pixels to work with, and good lighting on the face. Whether you can get that depends on the use case. For instance, when we worked with our national public broadcaster on capturing and deriving emotion and intention awareness in television viewers, things are fine in a 'laboratory' setting, but 'real living room conditions' where you get dark grainy low quality images from typical television 'Skype' cams were far more challenging.