Local directional intensity pattern for facial expression recognition

This paper presents a new face descriptor, local directional intensity pattern (LDIP), for human facial expression recognition. Considering the edge information and pixel intensity changes of face images, local features extracted from not only frequency domain but also spatial domain are encoded in a compact way, which makes features complementary to each other and much more discriminative. Additionally, the histogram refinement further reduces the feature dimensions while enhancing the performance of the texture descriptor. Extensive experiments are conducted to verify the effectiveness of our proposed method. The experimental results demonstrate that the LDIP is a discriminative and efficient facial expression descriptor, which achieves higher recognition rate than other descriptors.