Experimental evaluation of norm constraint sparsity exploitation for shallow water acoustic communication

Abstract

It has been recognized that the highly time-frequency selective underwater acoustic (UWA) channels exhibit sparse multipath structure, thus provides the possibility for performance enhancement in channel estimation and equalization. In order to exploit the sparsity contained in underwater acoustic channels, there have been increasing interests in the investigation of sparse estimation algorithms. From the viewpoint of practical engineering implementation, imposing different norm constraint on the least means square (LMS) adaptive iteration offers a convenient way for derivation of low complexity sparse channel estimation algorithm. However, while the practical underwater acoustic channels exhibit random time variations, there is a lack of experimental evaluation and comparison in terms of the performance enhancement of different norm constraint sparse channel estimation algorithms under practical UWA communication scenario, in a relatively long time-scale such as hours. This paper presents the experimental evaluation and comparison of some typical norm constraint least mean square (LMS) algorithm, i.e., the classic l0-norm, l1-norm and the non-uniform norm (NN), in the context of a shallow water acoustic communication system. Experimental results obtained in a practical shallow water channel within a duration of 5 hours are provided and discussed for evaluating the sparsity exploitation behavior in terms of communication performance achieved via a channel estimation based decision feedback equalizer. The relationship between the performance curves and ocean environmental observations like sound velocity profile, temperature field and wind speed is investigated to explain the comparison results.