D.S. Grigoryan

In problems of radar-tracking measurements in spatial and frequency area, a vital issue is possibility of angular coordinates or frequencies measurement not resolved on what or to parameters of secondary radiation sources in a low signal/noise ratio when signals sources are in limits a Rayleighan resolution element. The known autoregressional methods of short data spectral estimation don't allow to get the spectral estimates with the superresolution at threshold levels of signals in relation to own noise, characterized for radar problems. At low signal to noise ratio estimation the weight vector will be orthogonal to the signal-noise vector representing an additive mix of a signal and noise. One peak of spectral function on frequency of the seeming center of frequencies (angular coordinates) that doesn't allow to measure of each sources coordinates or frequency separately will thus be generated. Article purpose - to prove the criteria, allowing to measure frequencies of sources when there is no their superresolution on spectral function in autoregressive estimations. It is shown in work that at low a signal/noise ratio-s, when sources superresolution at estimation by a method of a linear prediction is absent, for solving of frequencies measurement problem it is possible to use criteria of a matched response minimum and an equivalent harmonic minimum, based on a minimum of a norm square of the narrow-band filter response on a signal from an exit of the linear prediction error filter. In article it is defined that for estimation of frequencies in the conditions of the superresolution absence of there is a necessity of transition to a class of incorrect problems and narrowings of area decisions. As set of decisions, from which it is offered a search method to choose the correct decisions, as much as possible corresponding to set criteria, it is offered to use set of pseudo-ideal weight vectors defined for all possible frequencies in limits of Rayleighan resolution element