Research Of Channel Estimation Ofr Wireless Communication System Based On Compressed Sensing

Along with the exploration of wireless communication technology and theexpansion of its application, the environment of wireless channel is becoming morecomplex. The quality of channel estimation directly affects the signal processing andinterpretation in receiver,and then the system performance. However, traditionalchannel estimation algorithms are usually ignoring the inherent sparsity in channelswith scattering space. The well-develop compressed sensing from signal collectionand processing, is even possible to recover original data from extraordinarilyincomplete measurements. Based on this frame and sparse representations forchannels, this thesis presents a more feasible and effective channel estimationalgorithm with overwhelming performances, which studies some key issue in sparsechannel estimation and optimal algorithms for overcomplete dictionary, bycombining the two steps of construction algorithms and channel estimationalgorithms. Both algorithm complexity and cost can be reduced by this technique, aswell as the overwhelming performance of channel estimation.In this thesis, compressed sensing is applied in estimating UWB channel withinherent sparsity, and we propose a filter matrix channel estimation algorithmleading to higher estimation quality than the random measurement estimationmethod. A module of finite impulse response filter is added in transmitter instead ofmeasurement matrix in receiver, which restrains the noise amplification as well asachieving dimension reduction of signal space. Furthermore, the thesis introducesthree reconstruction algorithms into the new channel estimation, including theDantzig Selector (DS), the Basis Pursuit De-noising (BPDN) and the OrthogonalMatching Pursuit (OMP). We analyze their advantages and show the effectivenessand opinions for choosing suitable reconstruction algorithms. In the thesis, the UWA channel estimation method for OFDM communicationbased on compressed sensing framework is studied, and a new approach to optimizedictionaries is proposed. We deal with the issues that lager mutual coherencebetween atoms in overcomplete dictionary will lead to less sparse channel, then tounsuccessful reconstruction, by adjusting the channel parameters to decrease themutual coherence in overcomplete dictionary. A fixed link between the averagemutual coherence of atoms and the factors: operating bandwidth, the number of pilotsubcarriers, and coherence bandwidth is analyzed, then the high-qualityreconstruction with lower complexity is achieved by designing the parameters as thenumber of pilot subcarriers. And the pilot information is collected to compose themeasurement matrix. In the thesis, the three-dimensional simulation figure is plottedto analyze the parameter conditions, and the optimized algorithm is translated in unitdisk. Furthermore, we employ the DS, the BPDN algorithm and the OMP algorithmsrespectively to illustrate the superiority of the optimized dictionary.According to the characteristics that the wireless communication for OFDM isextremely sensitive, and then affected by Doppler shifts in UWA channel, wedevelop the direct estimation and compensation technique based on compressedsensing with Doppler shifts. We introduce the new channel estimation method withno compensation for Doppler shifts, which releases the sparse channel estimationfrom the highly dependence on the accuracy of the estimation and compensation forDoppler shifts. We model the Doppler shifts as the atoms location shifts and sparseattenuation change in overcomplete dictionary, then the channel parameters can beestimated directly in the compressed sensing framework. The simulation resultsprove that direct CS-based channel estimator developed is capable of improving theestimation performance which performs better than estimators with compensationerror, although a little weaker than that with perfect compensation. This new channel estimators based on CS frame for OFDM UWAcommunication is extended to MIMO-OFDM system. The new channel estimationalgorithm based on CS for MIMO-OFDM system performs higher estimationaccuracy for making the most of sparse channel. The estimating sparse signals istaken as the impulse response of the channel between every receiving antenna and alltransmitting antenna, and then a series of non-orthogonal basis in a completedictionary are used to describe the signal under reconstruction. The effectiveness ofour algorithm is demonstrated by comparing the channel estimation performance ofBPDN, Dantzig selector and OMP algorithms. The experimental results show thatsparse channel estimation algorithm based on CS is superior to the traditional leastsquare algorithm of channel estimation in terms of accuracy, and can still estimatechannel parameters accurately, even in the case of least square matrix inversesingular.