Network with symmetrical neurons transformation function based on higher-order statistics for lowering the noise of unstationary signal

The method for network designing with symmetrical neurons transformation functions (SNTF) based on higher-order statistics (HOS) was analyzed. This network was shown to provide noise lowering for un-stationary signal. In the information-measurement systems noises with Gauss distribution and noises with non-Gauss but symmetric distribution are well suppressed. The structure of the network based on HOS is shown, and is compared with the network structure (having SNTF) and common learning LMS algorithm.