Personal Verification with Hand Shapes Using a Modular-Type Neural Network with RBF Output Units
There are increasing needs of personal verification technologies using physical characteristics of humans for automatic gate control systems. In this paper, we propose a personal verification system with hand shapes using a modular-type neural network with RBF output units. We extracted many features from a hand image with a simple algorithm, and then selected 20 features among them based on the results of the statistical analysis. These features are lengths of fingers and the area of a palm, etc. We used a set of the selected features as an input pattern to the modular-type neural network. Each module of the modular-type neural network is a three-layered neural network that has one RBF output unit. The modular-type neural network with RBF output units can achieve high rejection rates on patterns of unlearned classes. This is its advantage over the conventional modular-type neural networks with sigmoidal output units. We show that our system achieves both high verification rates on patterns of 40 learned persons and high rejection rates on patterns of 20 unlearned persons.
KeywordsMiddle Finger Cross Point Image Scanner Hand Shape Hand Image
Unable to display preview. Download preview PDF.
- 1.Nagano T. and Hirai Y. (1998): Personal verification with palm print and hand shape by utilizing neural network techniques. International ICSC/IFAC Symposium on Neural Computation, Vienna, Austria, pp. 398 - 404Google Scholar
- 2.Nagano T., Ishihara S. and Eguchi H. (1999): Hand shape has sufficient information for personal verification. 7th European Congress on Intelligent Techniques & Soft Computing, Aachen, Germany, pp. 271Google Scholar
- 3.Wegstein J. H. (1982): An automatic fingerprint identification system. NBS Special Publication, pp. 500 - 589Google Scholar
- 4.Rosen J. (1990): Biometric system opens the door. Mechanical Engineering, vol. 112, no. 11, pp. 58 - 60Google Scholar
- 5.Ishihara S. and Nagano T. (1999): A modular neural network with RBF output units. 1999 IEEE International Conference on Systems, Man, and Cybernetics, Tokyo, Japan, vol. 2, pp. V-344-349Google Scholar