Smile Expression Classification Using the Improved BIF Feature

Abstract

Biologically Inspired Feature is one of efficient feature descriptions, and achieves great performance in some applications. This paper proposes an improved Biologically Inspired Feature(IBIF), and applies this feature into smile recognition. The main contributions of our paper are as follows. 1) a rotation-invariant BIF feature is proposed, which adjusts the RBF function of the traditional Biologically Inspired Model(BIM), 2) the sparse coding method is introduced, and is to establish the Patch dictionary for changing the random patch selection of BIM. Some comparative experiments are made between IBIF and some popular features, such as Gabor, PHOG and BIF. The final experimental results reveal that the IBIF feature can achieve better performance, and can be efficiently applied into the real smile recognition system.