Lattice RLS for Nonstationary Signal Processing

Abstract

The paper presents the modification of lattice RLS adaptive filtering algorithms for the case of nonstationary signal processing. The modification includes the using of sliding window and dynamic regularization in the adaptive filter correlation matrix estimation. The algorithms can be implemented by means of sequential or parallel computations. Based on sequential computations, 30 regularized prewindowed, sliding widow and regularized sliding window lattice RLS algorithms were developed. A family of the same algorithms with parallel computations includes 21 units. Lattice algorithms are developed for the application in single-channel adaptive filters with complex-valued weights. Some computation procedures of the algorithms are listed. The results of the algorithm simulation are also presented.