Zavediuk T. Associative processors of correlation pattern recognition

Methods and means of processing of signals based on associative processors with the use of special neural components that implement correlation processing in Hemming space were first developed in the thesis. Response of the neuron to harmonic signal was researched and initial formalization of this neuron's pulse flow was made. Method of square-pulse conversion of harmonic signals was first proposed and significant simplification of analysts of harmonic signals' recognition was achieved. Method of correlation processing and recognition of signals with associative stackable memory was first proposed. Method of transmitting of information based on system of residual classes was first proposed. Structure of the device which implements the method of square-pulse conversion of harmonic signals based on additional input squarer into the structure of the device was proposed. The developed microelectronic model of neuron's axon is characterized by reducing of structure, hardware and time complexity comparing with the known structural analogues. Special processor for recognition of harmonic signals based on model of axon and dynamic recurrent neuron was developed. The structure and circuitry of multichannel device for computing the sign of the correlation function and it’s temporal and hardware complexity was calculated. Associative special processor with neural components was embedded in the system of recognition of leap and short circuits in high-voltage power lines.