Sparse channel estimation for multicarrier underwater acoustic communication: from subspace methods to compressed sensing

  • Authors:

  • Christian R. Berger;Shengli Zhou;James C. Preisig;Peter Willett

  • Affiliations:

  • Department of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, PA and Department of Electrical and Computer Engineering, University of Connecticut, Storrs, CT;Department of Electrical and Computer Engineering, University of Connecticut, Storrs, CT;Department of Applied Ocean Physics and Engineering, Woods Hole Oceanographic Institution,Woods Hole, MA;Department of Electrical and Computer Engineering, University of Connecticut, Storrs, CT

  • Venue:
  • IEEE Transactions on Signal Processing
  • Year:
  • 2010

Quantified Score

Hi-index 35.68

Visualization

Abstract

In this paper, we investigate various channel estimators that exploit channel sparsity in the time and/or Doppler domain for a multicarrier underwater acoustic system. We use a path-based channel model, where the channel is described by a limited number of paths, each characterized by a delay, Doppler scale, and attenuation factor, and derive the exact inter-carrier-interference (ICI) pattern. For channels that have limited Doppler spread we show that subspace algorithms from the array processing literature, namely Root-MUSIC and ESPRIT, can be applied for channel estimation. For channels with Doppler spread, we adopt a compressed sensing approach, in form of Orthogonal Matching Pursuit (OMP) and Basis Pursuit (BP) algorithms, and utilize overcomplete dictionaries with an increased path delay resolution. Numerical simulation and experimental data of an OFDM block-by-block receiver are used to evaluate the proposed algorithms in comparison to the conventional least-squares (LS) channel estimator. We observe that subspace methods can tolerate small to moderate Doppler effects, and outperform the LS approach when the channel is indeed sparse. On the other hand, compressed sensing algorithms uniformly outperform the LS and subspace methods. Coupled with a channel equalizer mitigating ICI, the compressed sensing algorithms can effectively handle channels with significant Doppler spread.