Parallel Recognition of Different Classes of Patterns

摘要

The paper examines an approach to parallel recognition of noisy signals (images) of differentclasses during their simultaneous entry at the input of a recognition device as an additive mixture. The idea of the approach is that the synthesized device is originally parallel in terms of its operation. It is astructure of interconnected channels, each transparent to signals of its class and rejecting signals of all otherclasses. For this system, it is important to stick to the following condition. Standard signals of the alphabetshould be the basis. At the learning stage, the alphabet of standard signals is transformed so as to make thesesignals independent of information. In processing the mixture of recognized noisy signals, the problems oftheir resolution, display, and parameterization are also being solved.