Facial expression recognition using Local Directional Pattern (LDP)

A robust face descriptor is an essential component for a good facial expression recognition system. In this paper, we analyze the performance of a new feature descriptor, Local Directional Pattern (LDP), for the representation of facial expressions. LDP features are obtained by computing the edge response values in all eight directions at each pixel position and then a code is generated according to the relative magnitude´s strength. Thus each expression is represented as a distribution of LDP codes. Different machine learning techniques are compared using Cohn-Kanade facial expression database for classification. Extensive experiments explicate the superiority of the proposed LDP based descriptor over other existing well known descriptors.