The normalized TDM-LeRoux-Gueguen algorithm for multichannel adaptive filters

Abstract

A new algorithm for the calculation of time dependent coefficients of a multichannel autoregressive lattice filter, using the least squares method, is presented. One of the main advantages of this algorithm is the totally scalar realisation, without any matrix or vector operation The applied time-division-multiplex (TDM) principle and a square root normalisation leads to a highly symmetric lattice filter with guaranteed stability. From the pure order recursive construction of the algorithm follows the completely independence between the coefficient calculation and the covariance estimation. That means, there is no round off error propagation in time and higher order recursive windows are possible. The algorithm can be seen as a extension of the generalized LeRoux-Gueguen algorithm.<>