E.P. Tichonov, M.L. Fedotenkova

With perfection of digital electronic equipment and of methods for mathematical signal processing interest in au-tomation of medical diagnostics enlarges. The decision of this issue is important for recognition of conditions of the patient with the least time and cost expenses. This decision is based on methods and algorithms of identification of biomedical signals and digital sequences formed of them. Diagnostics on electrophysiological signals is connected with researches into electric brain activity in various stages of a sleep, in the course of anaesthesia, classification of infringements of a rhythm of heart, etc. Solving the problem of informative fragments formation, it is supposed that as a result of a priori stored knowledge and medical indicators fragments of electrophysiological signals for all considered cases are defined: without pathological changes or with. Fragmentary representation of signals is possible on the basis of their indications, therefore in the first place is searching for methods and algorithms for partition of fragments by means of which sharing of an initial signal into fragments is carried out. As a measure indications is offered to use time intervals under adaptive transformation of signals with given function and inaccuracy. The set of intervals for different functions of reconstruction under fixed inaccuracy forms attributes vector for fragmentation of signal. Polynomial approximation can be used as a function of reconstruction. The adaptive algorithm iteratively carries out search of an average time interval of digitization. In this work characteristics of adaptive algorithm of time intervals sampling are being analyzed. The analysis based on results, has shown connection of time intervals with a fractal.