Research On Pilot Based Sparse Channel Estimation For OFDM System

The property of time-variant multipath from broadband wireless channel distorts the signals which go through the channel and lead to signal amplitude deviation and phase shift. In order to obtain the exact original signal, detection must be done to the signals received at the receiver. In Orthogonal Frequency Division Multiplexing (OFDM) system, coherent detection and incoherent detection might be two ways to detect the signals that have passed the channel. Coherent detection has its advantage over the incoherent one, but channel state information is the prerequisite for the usage of this method. Channel state information is obtained via channel estimation, which therefore is the first step prior to coherent detection. The accuracy of channel estimation is significant in that it affects the reliability of coherent detection at the receiver, and that’s why it is of both theoretical and practical values to research for an efficient and reliable channel estimation technique. To speak. of, traditional channel estimation does not take much advantage of sparsity of wireless channel. That’s why this thesis presents the result of research on pilot-based sparse channel estimation for OFDM system. The main work done includes:(1) Analysis of channel estimation based upon Discrete Fourier Transform. It’s found that DFT-based channel estimation is compatible for sparse channel estimation, because sparsity is able to reduce the computational complexity involved. A study of exponential smoothing, one part of time series analysis, has helped to get a view of theoretical principles of how exponential smoothing suppresses the noise. An improved channel estimation method based upon DFT is proposed in this thesis after combining exponential smoothing and DFT in order to enhance the resistance to noise and increase the accuracy of channel estimation. The results through simulation have indicated that this improved method is indeed superior to the estimation only based on DFT, especially in the context of strong noise.(2) Comparison of traditional pilot-based channel estimation with the one based upon compressing sensing. The latter method demands less pilots so that resource utilization rate is higher than the traditional method. Greedy algorithm of compressing sensing is more easily calculated and implemented, but it has weakness in terms of anti-noise capacity. Therefore, in this thesis, I have proposed an improved method which combines exponential smoothing with Compressed Sampling Matching Pursuit (CoSaMP, part of greedy algorithm). The simulation done has also shown that the improved method is superior to CoSaMP in that it costs less computational complexity and is compatible for low-speed situation.