An RLS-LMS alogorithm for lossless audio coding

Abstract

This paper proposes a cascaded RLS-LMS algorithm to mitigate the slow convergence problem of the LMS algorithm for highly colored inputs. The proposed algorithm has been successfully used to construct a large-order adaptive FIR predictor for lossless audio coding to model the ample tonal and harmonic components of general audio signals. It is found in our experiments that the RLS-LMS predictor provides superior prediction gain performance compared with the conventional LMS predictor, resulting in a better overall compression performance.