Multi-input multi-output underwater communications over sparse and frequency modulated acoustic channels

Abstract

This paper addresses multi-input multi-output (MIMO) communications over sparse acoustic channels suffering from frequency modulations. An extension of the recently introduced SLIM algorithm, which stands for sparse learning via iterative minimization, is presented to estimate the sparse and frequency modulated acoustic channels. The extended algorithm is referred to as generalization of SLIM (GoSLIM). The sparseness is exploited through a hierarchical Bayesian model, and because GoSLIM is user parameter free, it is easy to use in practical applications. Moreover this paper considers channel equalization and symbol detection for various MIMO transmission schemes, including both space-time block coding and spatial multiplexing, under the challenging channel conditions. The effectiveness of the proposed approaches is demonstrated using in-water experimental measurements recently acquired during WHOI09 and ACOMM10 experiments.

On estimating sparse and frequency modulated channels for MIMO underwater acoustic communications

“…By combining (11), (12), and (13), and by taking the negative logarithm of the cost function, the optimization problem formulated in (13) becomes (14), which can be solved using a cyclic optimization approach: at each iteration, one of the parameter vectors h m , p m , η m , and f m is updated while keeping the other three fixed. In this way, a single difficult joint optimization problem is divided into four simpler separate subproblems.…”

“…Moreover, GoSLIM-V simultaneously estimates the CIRs among all of the M N transmitter and receiver pairs. GoSLIM-V is developed based on the following hierarchical Bayesian model, similarly to (11) and (12):…”

“…Previously in [11], at each GoSLIM iteration, we zero pad the sequence {z evaluated at each frequency grid point using (18) directly. In this way, the complexity reduces to O(d y d g ) per GoSLIM iteration.…”

“…The number of unknowns in the frequency dimension, as a consequence, is significantly reduced. Under this assumption, CIRs and the underlying Doppler frequency could be estimated in a separate manner [10], or in a joint manner by employing an extension of the sparse learning via iterative minimization (SLIM) algorithm (SLIM was originally developed for sparse ISI channels [7]), referred to as generalization of SLIM (GoSLIM) [11]. Like SLIM, GoSLIM addresses sparsity through a hierarchical Bayesian model, and because GoSLIM is user parameter free, it is easy to use in practical applications.…”

“…Like SLIM, GoSLIM addresses sparsity through a hierarchical Bayesian model, and because GoSLIM is user parameter free, it is easy to use in practical applications. It is demonstrated in [11] that GoSLIM outperforms the separate estimation algorithm proposed in [10] in terms of estimation accuracy and robustness against suboptimal training sequences.…”