SNIP: Smile—neutral facial display intensity predictor

Many affective disorders can influence the capacity of emotion expression and understanding. For instance, in the case of depression, the frequency and intensity of smiling is diminishing when the disease proceeds. Analysing these changes by a psychologist could bring information about the effectiveness of the prescribed medical treatment. Therefore, this work describes a system which automatically analyses video recordings from sessions with patients and determines the length and intensity of smile. The proposed system exploits a standard database to train the classifier, which is later used to distinguish between the smiling and neutral facial expression depicted in the video frames. The obtained information is presented graphically, indicating the class membership and temporal smile intensity. The performed experiments revealed that the classification accuracy exceeds 85%. Moreover, the case studies show, that the variations in smile magnitude enable better insight into the smile analysis.