Multichannel stabilized fast RLS algorithm for nonstationary signal processing

This paper presents the computational procedure of the most generalized version of the stabilized fast Recursive Least Squares (RLS) adaptive filtering algorithm. It is considered the algorithm for the adaptive filters with an unequal number of weights in channels. The using of the sliding window and dynamic regularization, when the correlation matrix is estimated, allows the algorithm to process nonstationary signals. The algorithm is presented in two forms: sequential and parallel. The simulation of the algorithm in an acoustic echo cancellation task demonstrates the algorithm efficiency.