Abstract
We construct the authentication system for fingerprint with multi-steps, which were 3-layered neural networks with a backpropagation learning algorithm. The authentication system consists of large classification with 1 step and small classification with 16 steps. As a result, the system is useful for the authentication system with a very large number of input patterns. From the results, we found that the authentication rate at each 1 step was 99.985% and that the authentication rates of the system increased with the number of steps and were estimated by the equation, {100-(0.015)**N}, where N is the number of steps. Therefore, the authentication rates of the system are controllable by the number of steps.
