Local sparse represenation for driver drowsiness expression recognition

This paper proposes a novel local sparse representation for driver drowsiness expression recognition. A discriminative local structured dictionary is learned from the non-overlapping sub-blocks of four kinds of facial Gabor images. Then, a novel local descriptor is presented by using sparse coefficients obtained by the learned dictionary to represent the entire human facial expression, and a simple linear SVM is used for final recognition. We demonstrate the performance of this proposed local sparse representation method on our driver drowsiness datasets.