Compressive Estimation Of Fast Time-varying Channels In OFDM Systems

In orthogonal frequency-division multiplexing(OFDM) systems, precise and real-time channel estimation is essential for data demodulation. Based on the Nyquist sampling theorem, the traditional pilot-aided estimation methods need overloaded pilot symbols to reconstruct the fast time-varying channels,which result in reduced bandwidth efficiency. Therefore, how to achieve a good trade-off between accurate estimation result and pilot ratio is one of the most important problems in the field of channel estimation. To address this problem, this thesis proposes a compressed sensing(CS)-based estimation method for fast time-varying channels in OFDM systems. The main content of this thesis can be summarized as follows:1. Using the delay-Doppler sparsity of the selective channel model in OFDM systems, this thesis cast the channel estimation as a problem of reconstructing sparse signals. In order to solve this reconstruction problem, a CS-based iterative method is proposed. On the basis of the Pre-estimation of the sensing matrix, this method operates in an iterative, decision-directed manner. Compared with the exsiting sparse channel estimation approaches, the proposed iterative method is more robust to the Doppler-induced inter-carrier interference(ICI) and thus achieves improved estimation performance for fast fading channels. Besides, the restricted isometry property(RIP) of the sensing matrix used by the proposed method is proved, verifying the feasibility and reliability of the proposed method.2. By simplifying the estimation steps of the proposed iterative method, this thesis also presents a sparse estimation method with lower complexity. Using an approximated model of the channel system, this simplified method eliminates the unknown elements in the sensing matrix and measurement equation, thereby removing the pre-estimation and significiantly simplifying the estimation steps. Moveover, the performance of the simplified method can be further improved by the usage of a specific pilot arrangement, which not only reduces the influence of the ICI on the approximated system model, but also guarantees the stable performance of the CS reconstruction. Compared with the iterative methd, the simplified one achieves degraded estimation performance with reduced computation complexity.3. The application of model-based compressed sensing(MCS) theory to the channel estimation is studied. By exploiting the group(or block) structure of the channel sparsity pattern, a MCS-based sparsity method is proposed. Compared with the CS-based approaches, the MCS-based estimation reduces the number of the pontential sparse samples by making full use of the additional structure of the channel sparsity pattern. This advantage enables the latter to achieve more precise performance with even lower pilot ratio than the former.4. A Bayesian compressed sensing(BCS)-based estimation method is designed for some special scenarios, where the prior statistical information of the sparse channels can be obtained. The proposd Bayesian estimation method utilizes Baysian matching pursuit(BMP) and non-Gaussian Baysian matching pursuit(NGBMP), respectively, which enable it to address the channels with Gaussian prior and non-Gaussian prior. Further more, in order to exploit the group-sparse structure of the selective channels, this thesis combines the MCS theory and the Bayesian recovery algorithms mentioned above, which yield so called group-GBMP and group-NGBMP algorithms. The usage of the two group-sparsity Bayesian recovery algorithms further improves the BCS-based estimation. The simulation results demonstrate that, for the conventional statistical channel models, the proposed BCS-based approach can make full use of the group sparsity and the statistical information of time-varying channels, thereby considerably improving the estimation performance.5. The sparse channel estimation for multiple-input multiple-output(MIMO)-OFDM systems is studied. By utilizing the joint sparsity of the multiple channels in MIMO-OFDM systems, this thesis cast the channel system as a joint sparsity model. Then, a distributed compressed sensing(DCS)-based estimation method is presented. Different from the existing sparsity methods for MIMO systems, the proposed DCS-based method reconstructs the multiple channels jointly, not individually. This advantage permits a reduction of both pilot ratio and computation complexity. Moreover, a group-sparsity version of the DCS recovery algorithm is also presented, which makes the jointly sparse channel estimation even more effective and efficient.