Investigation Of Compressed Sensing Based Sparse Channel Estimation And Pilot Optimization In OFDM Systems

Recently, there has been a growing interest in compressed sensing(CS) theory which can successfully reconstruct a sparse signal from fewer samples than those obtained by Nyquist rate. CS theory can be applied to sparse channel estimation in orthogonal frequency division multiplexing(OFDM) systems and multi-input multi-output orthogonal frequency division multiplexing(MIMO-OFDM) systems to reduce the number of pilot symbols. As to the problem of pilot placement, equidistant pilot placement is generally optimal in conventional channel estimation, which is, however, not true in CS-based channel estimation. So the optimal pilot allocation in CS-based channel estimation should be investigated specially. This dissertation focuses on the CS-based channel estimation in OFDM, non-contiguous orthogonal frequency division multiplexing(NC-OFDM) and MIMO-OFDM systems and deals with their corresponding optimization of pilot locations. The main contributions are as follows.1) The estimation of sparse channel in time-delay domain in OFDM system is modeled as a reconstruction problem of CS. Based on minimizing mutual coherence and modified mutual coherence of the measurement matrix in CS theory, two criteria of optimizing the pilot pattern for the CS-based channel estimation are proposed. Simulation results show that using the pilot pattern designed according to any of the proposed two optimization criterions gives a much better performance than using other pilot patterns in terms of the mean square error(MSE) of the channel estimate as well as the bit error rate(BER) of the system. The optimal pilot pattern designed by minimizing modified mutual coherence could obtain larger performance gains than that designed by minimizing mutual coherence.2) A new channel estimation based on CS in NC-OFDM system is proposed. As for NC-OFDM systems in cognitive radio context, the infrastructure of CS-based channel estimation, the design of pilot patterns and the algorithm of channel estimation are explored. Simulations show that, under many patterns of deactivated subcarriers, CS-based channel estimation obtains much better performance than existing NC-OFDM channel estimation methods in terms of the MSE of the channel estimate as well as the BER of the system.3) CS-based MIMO-OFDM channel estimation is fulfilled by decomposing the MIMO channel into several SISO channels which are then estimated by CS reconstruction algorithms. By minimizing the mutual coherence of the measurement matrix in CS theory, two pilot allocation methods for the CS-based channel estimation in MIMO-OFDM systems are proposed. Simulation results show that using the pilot patterns designed by the proposed two methods gives a much better performance than using other pilot patterns in terms of the MSE of the channel estimate as well as the BER of the system. Moreover, the optimal pilot patterns obtained by the proposed second method based on genetic algorithm and shift mechanism could offer a larger performance gain than those by the first method based on minimizing the largest element in the mutual coherence set possessed by pilot patterns for all multiple antenna ports.4) Distributed Compressed Sensing(DCS) theory has been employed in the estimation of sparse channels between all transmit-receive antenna pairs in MIMO-OFDM systems. Deterministic pilot allocation of MIMO-OFDM systems is considered to improve the performance of DCS-based channel estimation. By transforming the problem of DCS-based channel estimation to a problem of reconstructing block-sparse signals, a class of mutual coherence-related criteria is first proposed for optimizing pilot locations. By employing the proposed criteria, a genetic algorithm-based method of optimizing the pilot locations is then presented. Simulation results show that the DCS-based MIMO channel estimation with optimized pilot locations can improve the spectrum efficiency by nearly 36% and the bit error rate(BER) performance by 1.5dB, as compared with the least square(LS) channel estimation with equidistant pilot locations. Moreover, the DCS-based MIMO channel estimation yields a 4.7% improvement in spectrum efficiency under the same BER performance over the CS-based channel estimation.