Facial asymmetry quantification for expression invariant human identification

We investigate the effect of quantified statistical facial asymmetry as a biometric under expression variations. Our findings show that the facial asymmetry measures (AsymFaces) are computationally feasible, containing discriminative information and providing synergy when combined with Fisherface and Eigen-face methods on image data of two publically available face databases (Cohn-Kanade (T. Kanade et al., 1999) and Feret (P.J. Phillips et al., 1998)).