Nonlinear Processing of the Signal Information Parameters under the Influence of Additive Narrow-Band non-Gaussian Noise

It is shown that the existing algorithms of noise filtering are effective only in the case when the mismatch between the measured parameters and their estimates is so small that the posterior probability density of the information signal parameters can be represented in a series in powers of the indicated mismatch not higher than the second. We have analyzed algorithms that are optimal in the region of small mismatch and qua-si-optimal for any distortions between the measured parameters and their estimates. This analysis is valid for any signal-to-noise ratio at the meter input with additive narrow-band non-Gaussian noise affecting the signal. Optimal algorithms for nonlinear signal parameter processing in quadratures and on the envelope at small mismatch are proposed, as well as structural schemes implementing them. These algorithms and schemes are optimal by the criterion of maximum likelihood function. It is shown that the criterion of minimum of the generalized mean square error can be used for obtaining quasi-optimal algorithms when mismatching is arbitrary. The above algorithms can be implemented by two types of discriminators. The first type of discriminators structurally coincides with the discriminators optimal by the criterion of maximum likelihood function. It is shown that the second type discriminators provide a smaller value of the generalized root-mean-square error.