Research On Compressive Sensing Based Sparse Channel Estimation In OFDM Systems

In order to address the need of high-bandwidth and high-speed transmission innext generation mobile communication system, OFDM technique will be widelyemployed in the future communication system. However, since OFDM has a highrequirement of orthogonality amongsubcarriers, the phase noise and frequency shiftcasued by time and frequency selective channels would have a strongly negativeimpact on the performance of OFDM system. Therefore, to have a prior knowledge ofchannel state information at the OFDM receiver is the key to fight against the channelfading and achieve an accurate receival of signals.In order to solve the problem that the conventional two-dimensional interpolationmethods fail to make an accurate estimation of doubly-selective sparse channels, wepropose a sparse channel estimatin technique based on compressive sensing theory.The new method can fully exploit the sparsity of physical channels in thetime-frequency domain and transform the OFDM channel estimation model into asparse signal reconstruction problem. Finally, the exact estimation of sparse channel isfulfilled by using the BP algorithm. Simulation results show that the new method cansignificantly reduce the number of pilots and thus improve spectrum efficiency. It canmake an exact estimation of sparse channel while the conventional FFT-Linear andFFT-FFT two-dimensional joint interpolation methods even fail to work.Besides, we also propose a second-order differential based sparsity adaptivereconstruction algorithm. This novel algorithm is capable to achieve sparse signalreconstruction without the prior knowledge of signal sparsity, solving the bottleneckof conventional algorithms. The algorithm adopts the second-order differential of thecorrelation vector as a reference variable and uses a threshold strategy to identify thenumber of atoms involving in the measuring process. It is capable to estimate thesignal sparsity in an extremely short time. The simulation results demonstrate that thesecond-order differential algorithm can perform a real-time reconstruction whilemaintaining an excellent performance.