SIGNAL PROCESSING METHOD

摘要

computer engineering; neural networks. SUBSTANCE: method depends on splitting optical signal into two light fluxes of which one is passed through diffraction gratings; fluxes obtained in the process are optically multiplied and passed through multipass classifying neural network of Hopfield network type and input image is identified; second light flux is passed through diffraction grating of optical bonding matrix of perceptrontype single-pass neural network; diffraction grating choice is controlled by Hopfield network output signal; second light flux passed through diffraction grating is optically multiplied to generate output control signal for selecting neurons bonding matrix. EFFECT: enlarged functional capabilities. 3 dwg