| Summary: | 碩士 === 國立交通大學 === 電信工程研究所 === 100 === In recent years, single carrier frequency domain equalization (SC-FDE) has been widely considered in the research of wireless communications as an alternative to the orthogonal frequency division multiplexing. The symbol decision performance of SC-FDE relies on accurate channel estimation. Traditional channel estimation techniques for SC-FDE are based on the least-squares (LS) principle, which however does not take account of the sparse nature of wireless channels. In this thesis, we study compressive-sensing (CS) based sparse channel estimation for SC-FDE, in which the training system is described by a circulant matrix. We first characterize the probability that a random circulant matrix can satisfy the restricted isometry property (RIP). The result is seen to be tighter than existing solutions. By using the Dantizg selector for signal recovery, we propose a new optimal training pattern via minimization of the mutual incoherence (MI). Simulation results are used to illustrate the performance of the proposed scheme.
|
|---|
