Research On Compressed Sensing Based Sparse Channel Estimation In Wireless Communication Systems

In modern wireless communication systems, wireless signal propagated in thewireless channel will be affected by multi-path time delay spread, Doppler shift,angular spread and other characteristics of multi-path fading, leading to the receivedsignal even serious distortion. Therefore, the study which can accurately andefficiently obtain the channel state information (CSI) is particularly important tocorrectly and efficiently recover the transmitted signal. While for this feature wirelessmulti-path channel has inherent sparsity, compressed sensing (CS) techniques appliedin wireless channel estimation has become an extraordinary valuable researchdirection.As a novel sampling theorem, compressed sensing which break through thelimitation of the original Nyquist sampling theorem can efficiently reconstruct thesignal by virtue of a small sampling signal developed by the signal characteristics ofsparsity. Compressed sensing has received attention and research in the fields ofimage processing, data compression, optical, pattern recognition, wirelesscommunication, etc. Compared with the traditional linear channel estimation method,CS-based means exploit less number of pilots to gain the same channel estimationperformance, significantly improves the utilization of the spectrum.In this thesis, we firstly will briefly introduce the property of the wirelesschannel and compressed sensing and then specifically describe the basic principlesand characteristics, and several common recovery algorithms in CS. After that somesimulations will be conducted to compare between traditional channel estimationtechniques and CS-based in Multiple-input multiple-output orthogonal frequencydivision multiplexing (MIMO-OFDM) and Multiple-input multiple-outputnon-contiguous orthogonal frequency division multiplexing (MIMO NC-OFDM)systems, respectively. Theoretical simulations show that CS-based channel estimationtechniques not only greatly improve the spectrum efficiency, but can be well appliedin some limited situations in MIMO-OFDM systems, particularly in MIMO NC-OFDM systems which as a commonly used standard of the Cognitive Radio(CR). In particular, according to the Bayesian knowledge, it combines the CS theoryand research the Bayesian-based CS channel estimation algorithm-fast Bayesianmatching pursuit (FBMP) through the prior model selection and an approximateminimum mean squared error (MMSE) estimate of the parameter vector and achievebetter performance in the final chapter three.