Conference publications

Abstracts

XVII conference

Regularization in problems of signal separation and restoring

Tarabardin M.

The Samara State Railway University, 18, 1-st Bezimyanniy st., Samara, 443066, Russia

The model of signals generation in linear systems with k sources of signals, d receivers and discrete time assumes, that generated signals are distributed from a source to the receiver on channels, and the accepted signal is superposition of the deformed signals with k sources.

As division and restoration of signals we shall understand calculation of a vector of generated signals on a known vector of the measured signals and a matrix of impulse responses (IR) of channels.

For the decision of a problem of signal restoring and separation in real time in conditions of aprioristic uncertainty and non-stationarity linear system adaptive algorithm LMS with Tikhonov's regularization is offered. The parameters of multichannel adaptive inverse filter (MAIF) , it is necessary to pick up so that to minimize functional with Tikhonov's stabilizer.

Modelling in MATLAB environment of the considered algorithm of division and restoration of signals has shown reduction of the given error of the decision by 7-10 % in a steady mode of operation of system and on 40-60 % in a unstable mode.

• view as PDF