Assessment of facial wrinkles as a soft biometrics

This paper presents results on the assessment of facial wrinkles as a soft biometrics. Recently, several micro features such as moles, scars, freckles, etc. have been used in addition to more common facial features for face recognition. The discriminative power of facial wrinkles has not been evaluated. In this paper we present results of our experiments on evaluating the discriminative power of wrinkles in recognizing subjects. We treat a set of facial wrinkles from an image as a curve pattern and find similarity between curve patterns from two subjects. Several metrics based on Hausdorff distance and curve-to-curve correspondences are introduced to quantify the similarity. A simple bipartite graph matching algorithm is introduced to find correspondences between curves from two patterns. We present experiments on data sets using manually extracted and automatically detected wrinkles. The recognition rate for these data sets using only the binary forehead wrinkle curve patterns exceeds 65% at rank 1 and 90% at rank 4.