Compressive Sensing-Based Adaptive Sparse Multipath Channel Estimation

Abstract

Sparse multipath channel estimation has recently attracted significant attention due to the sparsity of the channel in broadband wireless communication. Many algorithms have been proposed for sparse multipath channel estimation. Among them, the least mean square (LMS) algorithm, based on adaptive filter, has attracted much attention due to its low complexity and high robustness. However, LMS is usually degraded by the long training signal, which needs large storage space. This paper proposes an improved method that transmits a circulating, short training signal, samples the received signal at a lower rate, and utilizes LMS with ℓ0-norm (ℓ0-LMS) to estimate the sparse multipath channel. This method can achieve high robustness in additive white Gaussian noise (AWGN), and reduce the sampling rate while needing small storage space for the training signal. Numerical simulations are provided to evaluate the performance of the proposed method.