FACIAL PAIN DETECTION USING MULTI-LINEAR SUBSPACE LEARNING

Pain identification and measurement is a very important and sensitive aspect in medical fields profoundly needed in diagnosis, treatment, drug testing and more. Standard methods of pain identification rely on patient feedback (verbally or through other means), however these methods can be inaccurate due to subjective patient self-assessments, and more importantly, can sometimes be impossible to achieve in cases of strokes or with infants, hence the need for automated methods that standardize and objectify this pain measurement. Previous works in this domain, mainly those which rely on facial expression for pain identification, have achieved varying measurement accuracy with some reaching a correct prediction accuracy as high as 95.5%. Our aim with project is to enhance this precision measure through use of an unexplored data pre-processing technique called Multi-linear Subspace Learning, with the use of a tensor oriented version of the PCA algorithm titled MPCA.