A Comparison of Hand-Geometry Recognition Methods Based on Low- and High-Level Features

This paper compares the performance of hand-geometry recognition based on high-level features and on low-level features. The difference between high- and low-level features is that the former are based on interpreting the biometric data, e.g. by locating a finger and measuring its dimensions, whereas the latter are not. The low-level features used here are landmarks on the contour of the hand. The high-level features are a standard set of geometrical features such as widths and lengths of fingers and angles, measured at preselected locations.