Asymmetric region Local Binary Pattern operator for person-dependent facial expression recognition

Automatic facial expression recognition analysis is a challenging area, which finds applications in human-computer interaction, human-robot interaction and online multimedia communication to name a few. From available Local Binary Pattern (LBP) variant operators, a single operator cannot take care of all properties such as scale, robustness and discriminative ability. It also does not have control on the length of the feature histogram obtained. This forces us to choose among LBP variant operators before their use in a particular application. This paper proposes an Asymmetric Region-Local Binary Pattern (AR-LBP) operator along with modified convolution technique for facial expression recognition. The AR-LBP operator has the properties of basic LBP, Extended-LBP, Multi-scale Block-LBP(MB-LBP) and mitigates the disadvantages of these operators, with respect to scale, discriminative ability of the operator and the length of the feature histogram. The operator proposed was evaluated for person-dependent facial expression recognition on JAFFE and FGNET databases using an multi-class SVM classifier with radial basis function as kernel. The maximum facial expression recognition rate was found to be 95.71% and 79.41% on the JAFFE and FGNET databases respectively.