Abstract:
Hand Geometry is one of the relatively new physiological biometrics used for
verification of a human being due to its stable and unique characteristics. There are
few open literatures addressing the research issues underlying hand geometry based
identity verification. The existing hand geometry based personal verification systems
need a lot of features such as lengths, heights, and widths of different fingers, so a
little bit of displacement of hand makes it difficult to acquire data non-intrusively.
The existing systems use pegs on their platfonn. If users give pressure then it is hard
to collect accurate data. All the methods need complex image acquisition systems.
This thesis presents a new biometric approach for the person verification based on
Hand Geometry. This work attempts to improve the performance of hand geometry
based person verification by an efficient method, which is simple, fast, easy to
handle and cost effective compared with other biometric person verification
techniques. The image acquisition setup used in this work is inherently simple, since
normal scanner is used to capture the hand images and it does not employ any special
illumination nor does it use any pegs to cause any inconvenience to the users.
Four methods -Joining Point Triangle (JPT), Pentagon, Tri Triangles (TT) and
Combined Approach are proposed. For features measur~ment, three or five points in
the hand image are considered. The JPT method is based on only three points.
Pentagon, TT and Combined methods are based on five points. Absolute, Weighted
Absolute, Euclidean and Weighted Euclidean distance metrics are applied to JPT,
TT, Pentagon, and Combined Approach methods for verification. Using the features
vectors of JPT, Pentagon, TT and Combined methods, Total Success Rates (TSR) are
98.99%, 99.19%, 99.60% and 99.78%, respectively. The remarkable achievement
obtained from the proposed method is the result of verification, which is the best
among the prevailing techniques of hand geometry based systems. Since the
proposed method reqUIres very small computational time, the practical
implementation is also easily possible.
