Application of artificial neural networks forhandwritten biometric images recognition
A Malygin1, N Seilova2, K Boskebeev3, Zh Alimseitova2
COMPUTER MODELLING & NEW TECHNOLOGIES 2017 21(1) 31-38
1Penza State University, Penza, Russia
2Kazakh National Technical Research University named after K.I. Satpaev, Almaty, Kazakhstan
3Kyrgyz State Technical University named after I. Razzakov, Bishkek, Kyrgyzstan
The development of information technology leads to new requirements for the development of security systems, identity authentication and other protection mechanisms. The article is devoted to the use of artificial neural networks for handwritten biometric images recognition that are used in high-authentication systems. There is given a general structure of the biometric-neural network authentication system, the structural scheme of information processing in biometric-neural network authentication systems, the structural scheme for learning the neural network converter of the biometric parameters vectors in the key code (password). There is formed and trained a network of neurons, are formed neural network containers on the basis of structures. The choice of the length of the biocode of neural network converters is substantiated. After graduation, testing is conducted and the probabilities of errors of the first and second kind are determined. There is given an example of a software implementation, where are given a learning mode, checking the results of training, testing the results of training.