Processing of two-dimensional data with use of two-rate Kohonen's neural networks

One of the attractive areas of mutirate phenomena research is that of image recognition. In this paper is offered the two-rate recognition of the English alphabet by means of the Kohonen's neural networks. The received two-rate Kohonen's neural network (TRKNN) not only is capable to distinguish the input images, but also is capable to restore images. Further analysis of obtained neural networks can be produced by means of the modern computer software. In example of this paper, the TRKNN is consisted of the "fast" and "slow" subsystems. The structure of the "fast" subsystem is simpler than the structure of the "slow" subsystem. The "fast" part of the TRKNN classifies an image, and the "slow" part of this neural network restores this image. The simulations for the TRKNN were carried out in a MATLAB/Simulink environment. This example shows the computing procedure and applicability of TRKNNs for fast-acting image recognition