Description: Determination of almost Gaussian signals is given, as one of types of semi continuous processes. On the basis of nonlinear task of minimization and method of variation calculation the decision, allowing to define transitional description of nonlinear filter - discriminator which carries out the functions of electoral filtration and estimation of value signal, is found. The given transitive characteristic takes into account aprioristic statistical data on an expected signal and a real level of capacity of a handicap. Advantage of the got result is shown on the indicator of the middle square of error, arrived at by comparison to the known linear method of filtration.
Keywords: nonlinear filtration, middle square of error, semicontinuous signals, function of verisimilitude