Abstract: By virtue of an increase in spectral efficiency by reducing the transmitted pilot tones, the compressed sensing (CS) has been widely applied to pilot-aided sparse channel estimation in orthogonal frequency division multiplexing (OFDM) systems. The researches usually assume that the channel is strictly sparse and formulate the channel estimation as a standard compressed sensing problem. However, such strictly sparse assumption does not hold true in non-sample-spaced multiple channels. The authors in this article proposed a new method of compressed sensing based channel estimation in which an over-complete dictionary with a finer delay grid is applied to construct a sparse representation of the non-sample-spaced multipath channels. With the proposed, the channel estimation was formulated as the model-based CS problem and a modified model-based compressed sampling matching pursuit (CoSaMP) algorithm was applied to reconstruct the discrete-time channel impulse response (CIR). Simulation indicates that the new method proposed here outperforms the traditional standard CS-based methods in terms of mean square error (MSE) and bit error rate (BER).

Key words: channel estimation, model-based compressed sensing, non-sample-spaced multipath channels