The noise influence of hardware and environment on signals in information channel reduces the recognition accuracy of patterns, represented by these signals. The main methods of noise influence elimination on recognition accuracy are: the selection of invariant for noise features and also by filtration. Usage of the second method very much often results to unjustifiable computing complexity, as the efficiency of filtering for given pattern arrangement in feature space is not considered. That is why the application in many recognition tasks of not complicated filters can increase the recognition accuracy in the same measure, as filters with considerable computing complexity (for example, Kalman filters). In this paper, the simple method of a fast signals filtration which minimizes computing complexity is reviewed.