Separate Analysis of Informational Signs in Multi-Parametric Combined Patterns Recognition Systems

Abstract

The objects for analysis in multi-parametric combined patterns recognition systems are repre-sented by several patterns with different nature of origin. This allows to increase the number of informative signs and the classification reliability. However, there is a negative aspect – increas-ing the time complexity of the data analyze. The aim of the study is reducing the time complexi-ty of data analysis and decision making in multi-parametric combined pattern recognition sys-tems. Well known solutions to this problem: data processing algorithms complication, boosting algorithms using and install more productive computing systems. However, more promising is use an approach that takes into account signs informativity level and decision-made occur does not across the entire data set. For achieve the aim of the study is being solved important scientific problem – developing the new method separate analysis of information signs in multi-parametric combined patterns recognition systems. This method is based on the fact, that in the multi-parametric combined patterns recognition systems the object is represented by a set of patterns which have a different nature of origin. Therefore, in the case of correct classification, all pat-terns from the totality allow to relate the object to the same class. The patterns of recognition object are compared until one or several of them is made a decision on the classification. In such an approach a full analysis of all signs is not carried out, which reduces the time complexi-ty of the recognition process and, as a result, the classification decision-making process speed up.

Keywords

• recognition
• a
• classification
• patterns