Wireless Communication Signal Processing Methods Based On Compressed Sensing

With the development of wireless communication technologies, thebandwidth of a signal included in the signal processing is increasing dramatically,which makes the traditional Nyquist theorem-based signal processing methodshave higher and higher demand for the sampling rate. However, the too highsampling rate requirement will lead to a severe challenge of the conventionalhardwares and softwares. Therefore, how to effectively realize the broadbandsignal processing under low sampling rate conditions has become a hotspot ofcurrent wireless communication research.In recent years, the emerging theory of compressed sensing (CS) makes thereconstruction of sparse or compressible signals from low-dimensionalobservation signals possible. Applying the CS theory to process wirelessbroadband sparse signals can overcome the Nyquist theorem restriction, andsignificantly improve the usefulness of the signal processing methods. Hence, thisthesis focuses on the CS-based sparse channel estimation, the CS-based impulseradio ultra-wideband (IR-UWB) signal detection, and the CS-based orthogonalfrequency division multiplexing (OFDM) system’s impulsive noise interferencemitigation. The main contributions of this thesis are the following.1. For the sparse channel estimation problem in two-way relay network, animproved orthogonal matching pursuit (IOMP) estimation algorithm is proposed.The IOMP algorithm uses an iteratively reweighted least squares estimation toreplace the least squares estimation process of the existing orthogonal matchingpursuit algorithm, gradually reduces the impact of outliers on the estimates, thusimproving the estimation accuracy of sparse two-way relay channel.2. For the sparse channel estimation problem in multiple input multipleoutput (MIMO) system, a modified compressive sampling matching pursuit(MCoSaMP) estimation algorithm is proposed. The MCoSaMP algorithm firstadaptively selects several elements as an index set, then uses a backtrackingstrategy to delete some incorrect elements in the index set, and finally estimateschannel according to the updated index set. This new algorithm not only avoids the requirement of known sparsity, but also shows a good trade-off betweenestimation performance and computational complexity.3. For the IR-UWB signal detection problem, a CS-based IR-UWB signaldetection system is designed by exploiting the sparsity of the time domain IR-UWB signal, and an adaptive correction matching pursuit (ACMP) signaldetection algorithm is proposed. The ACMP algorithm first deduces the sparsityexpression by using the specific parameters of IR-UWB transmitter, then avoidsreselecting the optimal vectors chosen in the previous processing via anorthogonal process, and finally uses an adaptive correction factor to ensure therobustness of the algorithm, thus improving the success probability of detection.4. For the OFDM system’s impulsive noise interference mitigation problem, aCS-based mitigation system model is designed by using the impulsive noise’sprojection onto null subcarriers or pilot tones in OFDM symbol. Based on theabove system model, a space alternating sparse Bayesian learning (SASBL)algorithm and a subproblem approximation algorithm are proposed respectively.The SASBL algorithm first estimates the hyperparameters through alternativelyupdating iteration, and then solves the posterior mean from an equivalentexpression, thus improving the performance and reducing the computationalcomplexity. The subproblem approximation algorithm first transforms theimpulsive noise estimation problem into an l2－l1problem, then achieves itsoptimum solution by iteratively solving an optimization subproblem, and finallycancels the impulsive noise interference by subtracting out the estimate. Thisalgorithm can significantly reduce the computational complexity without muchperformance loss.