Abstract
The question of perception attracts a lot of investigation concerning fundamental issues of knowledge representation and particularly application-oriented techniques. We clearly indicate the context-dependent character of information processed in cognitive processes (as opposed to statistical analysis in which no context is taken into account). The problem being addressed here is translated into an optimization scheme under which algorithms of data analysis are worked out. A general class of techniques dedicated to linguistic analysis of records of numerical data is introduced. Its essence relies on determining the level of data and fuzzy labels.