Definition of weight coefficient in two-stage automatic compensation based on LMS-algorithm and interference correlation characteristics

Abstract

Introduction. Two-stage digital automatic compensators of active noise interference based on LS and RLS-algorithms provide parallel processing of signals and capability to connect/disconnect additional compensation channel block. LMS-algorithm is widely used for synthesis of digital automatic compensators of active noise interference. Its advantage is low computational cost. When solving certain practical problems correlation characteristics of interference signals are either given or their estimates are obtained. Problem statement. The first stage of the two-stage digital automatic compensator provides compensation of interference in the main channel using compensation channel signals of one of the blocks and orthogonalization of signals in the other block. The second stage provides compensation of the residue interference using the obtained orthogonalized signals. Modules of the first and the second stage are single type weight totalizers. We need to determine weight coeeficient of the two-stage digital automatic compensator with orthogonalization of signals of part of compensation channels based on LMS-algorithms and interference correlation characteristics. Using LMS-algorithm there was obtained two-stage digital automatic compensator with orthogonalization of part of compensation channels that is quasi-optimal regarding the criterion of minimum mean square error and it’s feature is lower computational complexity compared to two-stage digital automatic compensators based on LS and RLS-algorithms. It can operate in modes when preocessing of the complete training package of the input data is carried out in the first and then in the second stage as well as in the mode of simultaneous operation of the stages. There have been acheived analytical expressions for defining weight coefficients of modules in the first and in the second stage of the automatic compensator determined on the basis of selective and precisely known correlation characteristics of input interference signals. LMS-method based two-stage digital automatic compensator efficiency analysis with simultaneously operating stages and its comparison with a known one-stage automatic compensator with the same number of compensation channels were conducted with the help of static modeling at various conditions of interference correlation matrix and various numbers of interference sources. Conclusions. The obtained two-stage digital automatic compensator with orthogonalization of part of compensation channels is quasi-optimal regarding the criterion of minimum mean square error and it has lower computational complexity compared to two-stage digital automatic compensators based on LS and RLS-algorithms. Weight coefficients of the automatic compensators determined on the basis of selective and precisely known correlation characteristics of input interference signals are optimal regarding the criterion of least squares and minimum mean square error respectively. The obtained two-stage digital automatic compensators provide capability to connect/disconnect additional compensation channel block and appropriate computational stage modules proportionately to the interference situation.