Abstract
There is very limited literature currently on the use of Sparse Representation (SRC) for the recognition of facial expressions and as far most facial expression analyses; they have been based on posed image databases. These comprise of expressions that often differ from the realistic displays of the expressions that depict affective states. To offer a more practical solution, we apply a recently proposed approach for SRC to the Facial Expression Recognition (FER) problem using the recently developed Natural Visible and Infrared facial Expression (NVIE) database of spontaneous images. We expand the database in order to satisfy the condition of an underdetermined (overcomplete) dictionary and present results showing better recognition rates for spontaneous images than in the existing literature (albeit limited).