Research On Sparse Channel Estimation For Underwater Acoustic Communication System

Acoustic wave has good propagation characteristics in underwater environment and is the most widely used underwater communication mode at present.However,there are also problems such as limited available bandwidth,large propagation delay and multipath fading.Orthogonal Frequency Division Multiplexing(OFDM)is a kind of parallel multi-carrier modulation technology.It divides the channel into several orthogonality sub-channels for information transmission,which improves the transmission rate and Frequency utilization.It has become one of the research hotspots of high-speed underwater acoustic data transmission.The channel estimation technology can improve the accuracy of data demodulation by estimating the channel state information.However,due to the complexity of underwater acoustic channels,traditional channel estimation techniques require large quantum carriers to transmit pilot information,resulting in serious waste of spectrum resources.Because of the sparsity of underwater acoustic channel,the channel parameters can be estimated by a small amount of pilot data in Compressing Sensing(CS),so as to reduce pilot cost and improve spectrum utilization.Reconstructing algorithm and pilot structure design are two key problems in compressed sensing channel estimation,which directly affect the performance of channel estimation.In this paper,the sparse channel estimation method of OFDM underwater acoustic communication system is studied in depth for these two problems.The main work is as follows:1.Multipath Matching Pursuit(MMP)algorithm has better estimation accuracy in channel estimation,but it has high computational complexity and requires a priori information of channel sparsity.To solve this problem,this paper proposes a Cross Validation(CV)combined with Regularization,Multipath Matching Pursuit Based on Cross Validation and Regularization(CV-RMMP)algorithm for OFDM underwater acoustic channel estimation.CV provides a stop criterion for the algorithm,so that the algorithm can reconstruct the signal without a priori information of channel sparsity or noise level,and check whether the algorithm is over-fitting,which improves the accuracy of the channel estimation.Regularization is used to further filter candidate sets and reduce computational complexity and storage overhead.The simulation results show that,compared with the original MMP algorithm,the proposed algorithm reduces the time complexity by 90% without reducing the estimation accuracy,and has better estimation performance,and does not need the priori information such as channel sparsity,so it has a high practical value.2.Restricted isometric Property(RIP)of measurement matrix is required for channel estimation based on compressed sensing,which cannot be satisfied by traditional isometric pilot distribution method.Stochastic Sequential Search(SSS)algorithm is based on the minimum cross-correlation criterion to Search the optimal pilot structure through internal and external double cycles.However,this algorithm only selects a minimum value for each iteration,and the selected optimal pilot position is easy to fall into the local optimal solution,resulting in poor convergence.Aiming at this problem,in this paper,the compressed sensing channel estimation sprocket frequency structure design problems were studied,this paper proposes a optimization based on adjustment of cross-correlation criterion pilot search algorithm,this algorithm within each cycle when select more than one possible pilot episode,increases the circulation within the search scope,improve the probability of the optimal pilot location is selected.Compared with SSS algorithm,it has higher stability.Simulation results show that the proposed pilot setting optimization scheme improves the accuracy of channel estimation,and saves 4d B SNR compared with SSS algorithm at the same 1% bit error rate.