The method of estimation of extreme values of nonstationary random signals dispersion

The method of extreme values of dispersion's estimation of nonstationary random signals, as a linear dynamic objects (LDO) reaction caused by switching on a random stationary input with definite dispersion and normalized correlation fonction is suggested. The method uses some features of LDO reaction caused by a determined input with limited absolute range. There are two kinds of estimations obtained - for LDO with constant sign and nonconstant sign pulse response characteristic (PRC). For LDO with constant sign PRC the extreme value estimation is proportional to the squares of unit-step response. For LDO with nonconstant sign PRC the extreme value estimation contains the sum of integrals with coefficients. Each integral is calculated on interval where PRC has constant sign. Each coefficient is a sum of integrals of PRC's modules or integrals of PRC's modules calculated on intervals where PRC has constant sign. The obtained estimations has a form of monotonously growing fonctions and can be simply calculated.