Research On Compressed Sensing Based Channel Estimation Method

In wireless communication system, the accurate and real-time channel estimation is necessary for data demodulation at the receiver. Traditional linear channel estimation methods usually assume that the wireless channels are rich multipath and use a large number of pilot tones to obtain accurate CSI, leading to low spectrum efficiency. Compressed sensing, as a new kind of data sampling and processing theory, is able to recover the original signal accurately from a small amount of measurements, which opens up a new research direction for channel estimation. In this thesis, based on the sparse characteristic inherent in wireless multipath channels, three improved algorithms are presented according to certain conditions in sparse channel reconstruction. The main contributions are summarized as follows:As the classic compressed sensing reconstruction algorithms perform poorly in the accuracy of channel estimation, we propose an improved algorithm based on discriminant analysis in this thesis. Introducing the idea of classification, the proposed algorithm distinguishes the real propagation path and the noise by calculating the Mahalanobis distance between the taps, which makes it possible to restore the channel state information in noisy measurement. Simulation results show that the proposed algorithm performs better than the classic reconstruction algorithms in all SNR range.For most of the compressed sensing reconstruction algorithms require the sparsity as a priori knowledge in channel estimation, we analyze a sparsity adaptive algorithm—SAMP in this thesis. Considering the slow convergence speed of SAMP algorithm and the computational efficiency requirements of channel estimation, we propose an improved adaptive algorithm based on correlation analysis in this thesis. The algorithm estimates the channel sparsity all at once by the correlation analysis and then applies it to the backtracking iterative reconstruction. In this way, the channel estimation could be done under the minimum number of iterations. Compared with the SAMP algorithm, the proposed algorithm achieves approximate performance but with lower computing complexity.Considering that the existing compressed sensing based channel estimation methods only focus on the channel estimation in particular moment, we introduce the dynamic compressed sensing into channel estimation in this thesis. Utilizing the sparse characteristics and time-vary ing characteristics of wireless channel, the proposed method establishes a down sampling state variable equation based on the sparse multipath, and then it performs a reduced order Kalman filter on the multipath to obtain the estimation of channel state information. Simulation results show that the proposed method improves the efficiency and accuracy comparing with the traditional deterministic compressed sensing method.In conclusion, focusing on the sparse channel reconstruction, we put forward three corresponding improved algorithms on the basis of existing research. The methods proposed in this paper are useful to overcome the problem in the application of compressed sensing based channel estimation.