Comparison of Sparsity-Aware LMS Adaptive Equalization for Underwater Acoustic Communications

Abstract

The least mean squares (LMS) type direct adaptive equalization (DAE) has been a favored choice in underwater acoustic communications. As the equalizer size grows, however, conventional LMS DAEs face serious challenges in both performance and complexity, which has motivated the development of the sparsity-aware alternates. By taking advantage of the sparse (non-uniform) structure of the equalizer, the so-developed DAE achieves performance improvement and/or complexity reduction. There are two families of sparsity-aware DAEs in the literature, based on the proportionate-updating (PU) principle and the zero-attracting (ZA) theory, respectively. Even though, there is no consensus on which family is a preferred choice. In this paper, the improved proportionate normalized LMS (IPNLMS) DAE designed with the PU principle and the selective ZA (SZA) NLMS DAE based on the ZA theory are compared, aiming to gain preliminary insights on the choice of the sparsity-aware DAE for practical applications. Analytical studies show the two have comparable complexity, and experimental results indicate the IPNLMS-DAE is a better choice than the SZA-NLMS-DAE, in terms of performance and maneuverability.