Blind Identification Algorithms for Sparse SIMO Systems
Abstract
In this work, we are interested in blind identification of sparse single-input multpe-output (SIMO) systems. First, we present a modified cross-relations (CR) technique combined with a $\ell_p$ norm, which is considered as a good sparsity measure. Then, A maximum a posteriori approach is considered using generalized Laplacian distribution for the channel coefficients. This leads to a cost function given by the deterministic maximum likelihood (ML) criterion penalized by 'a sparsity measure' term expressed by the $\ell_p$ norm of the channel coefficient vector. A simple but efficient optimization algorithm using gradient technique with optimal step-size is proposed. The simulations show that the proposed method outperforms the ML technique in terms of estimation error and is robust against channel order overestimation errors.
Origin : Files produced by the author(s)
Loading...