Fevralev D. Combined bispectrum filtering technique for signal shape estimation at the influence of high-intensity noise

Object – signals represented as a set of realizations with slightly varying or constant shape from one realization to another that are formed by systems of primary processing of radar signals, passive systems of target tracking, radio device sensors, other measuring systems. The goal is to increase ac-curacy of signal shape recovery using a set of mutually shifted realizations in conditions of low signal-to-noise ratio and lim-ited a priori information on received signal and present noise by using combined methods of bispectral analysis. Methods – probability theory and mathematical statistics, numerical simulations, digital non-adaptive and adaptive nonlinear filter-ing, estimation theory. Novelty – first time statistical charac-teristics of interference present in bispectrum of noisy signal and in Fourier spectrum recovered from bispectrum are stud-ied; new approaches to smoothing bispectrum estimate and Fourier spectrum recovered from bispectrum by discrete co-sine transform based filtering are proposed; ways to combin-ing robust estimates and nonlinear adaptive filtering for im-proving bispectrum estimates and signal shape recovery are improved; adaptive DCT filters based on local estimation of variance for smoothing 1D and 2D processes corrupted by non-gaussian non-stationary noise are designed; first time a two-stage method for signal shape recovery from an ensemble of mutually shifted realization based on combining bispec-trum-filtering and correlation methods is proposed. The pro-posed methods are realized as software-algorithmic means and implemented in several enterprises. Application area – recov-ery of signal shape for small signal-to-noise ratios at secon-dary processing stage.