BUILDING OF RECOGNITION OPERATORS IN CONDITION OF FEATURES’ CORRELATIONS

The problem of recognizing of patterns given in the space of correlated features is considered. The new approach to the building of modelof recognition operators, which considers the correlation of given features, is proposed. The building of the model is carried out for potentialfunction type recognition operators. The main idea of the proposed approach is formation of uncorrelated subsets of strongly correlatedfeatures and extracting preferred correlation model for each of subsets of strongly correlated features. Analysis of the results shows that theconsidered recognition operators are used in cases when there is a certain correlation between objects belonging to the same class. When theexpression of this relationship is weak, classical model of recognition operators is used. The main advantage of the proposed recognitionoperators is to improve the accuracy and the significant reduction in the volume of computational operations in recognition of unknownobjects, which allows them to use when building recognition systems working in real time.

Hali tarjima qilinmagan