Fake smile detection using linear support vector machine

Fake smile is an emotional sign on the face that can be used as information for non-verbal communication. One of its functions is for lie detection purpose based on the information of emotional sign generated on the face. The emergence of fake smile indicates that there are negative emotions, uncomfortable feeling, and something hidden in a person. This research aims to detect fake smile. In fact, real smile is characterized by the contraction of zygomatic major muscle on the edge of mouth corner and obicularis oculli muscle on the eyelids. However, on a fake smile, zygomatic major muscle experiences contraction, but obicularis oculli muscle doesn´t contract. Contraction of the zygomatic major muscle is identified by the appearance of wrinkles on the cheeks corner of the mouth, whereas obicularis oculli contraction is identified by the feature value of eye elongation. On the test image, segmentation of RoI (Region of Interest) is done on cheeks and eyes. On the RoI (Region of Interest) cheeks, wrinkle density is calculated; whereas elongation value is calculated on the RoI (Region of Interest) eyes. Based on the two variables above, with support vector machine linear for its classification, smile is classified into two classes, i.e. real smile and fake smile. The test result showed that the accuracy of system is 86 %, whereas the error rate is 14%.