Shape-based hand recognition

In this thesis, a new algorithm for personal recognition has been proposed.Personal recognition has been made using geometric features of the hand of theindividuals. Hand geometry is a biometric that identifies individuals by the shape oftheir hands. In this study, a smart, simple and an effective algorithm is presented forboth verification and identification of the people.The proposed algorithm consists of three major parts: Segmentation forobtaining the hand contours, feature extraction and recognition. Segmentation partseparates hand region from the background and extracts the hand and fingercontours. Feature extractor finds the lengths of the fingers, finger widths at differentheights of the fingers, and the width of the palm as a feature vector. Finallyrecognition part compares the feature vector obtained from the input image with thedatabase and gives the result for identification or verification purposes.This algorithm uses right hand images of the individuals for both verificationand identification purposes. The necessary codes for the proposed algorithm werewritten in Matlab software. To see the overall performance, the proposed algorithmhas been tested over a large number of images. It was obtained that the identificationrate for this algorithm is 97.44 % and verification rate is 98.72 % giving the leasterror.Key Words: biometrics, segmentation, hand geometry, smearing, templatematching, recognition