Abstract:
One of the goals of biometric systems is to identify a person automatically based on
his/her biometric characteristics. These may include iris, fingerprint, face, voice, or
hand. Hand geometry recognition is one of the biometric characteristics that can be
used to distinguish a number of individuals, because everyone has different hand
lines, shapes and sizes. This project study involved use of trained hand images for
personal verification and identification to login and access a personal computer
system based on windows operating system. This was to add on general system
security to reduce vulnerability in password attacks for example remote key
loggers. Hand geometry features used in this study consists of the lengths of fingers
from the centroid of the palm image, and Euclidean distance between hand finger
tips and valleys (25 features). The features were used in the thresholding phase, and
Neural Networks classification was used in the training phase of the hand shape.
Phases were concatenated and was used to determine the identity of the person.
Users can place their hands freely on the glass panel inside the image acquisition
device, and hand images were acquired using a Kinect camera infrared rays to
generate hand edged image which was used for recognition. Project test datasets
included 10 users. These were trained and used for testing system parameters.
Accuracy obtained was 70%. False Rejection Rate (FRR) returned 10%, and False
Acceptance Rate (FAR) 20% and Genuine Acceptance Rate (GAR) was 80%.
