Research On Sparse Bayesian Learning Based Sparse Channel Estimation In Underwater Acoustic OFDM Communication

Underwater acoustic(UWA)communication technology is an important technical which is widely used for exploring,developing and utilizing marine resources.However,the UWA channel is one of the most challenging wireless communication channels with large delay spread,significant Doppler effects,frequency-selective fading,seriously limited bandwidth and other significant characteristics.And the channels are time-varying and space-varying.With the increase of marine activities,the traditional low-rate underwater acoustic communication technology has been unable to meet the requirements of communication performance.It is also an urgent research direction to study the physical layer communication technology with high communication rate and high frequency band utilization.So channel estimation issue in underwater acoustic communication is studied here,based on the orthogonal frequency-division multiplexing(OFDM)technology which has the characteristics of resisting multipath channel fading and high spectral efficiency.In this paper,the characteristics and models of underwater acoustic channels are studied theoretically and analyzed with experimental data.Secondly,without considering the prior information of the sparse channel,the sparse channel estimation method based on compressed sensing(CS)is studied and an improved algorithm is proposed to reduce the computational complexity.Then under the premise of considering the prior information of sparse channel,the sparse Bayesian Learning(SBL)theory is studied,the time correlation based sparse Bayesian learning channel estimation and the SBL channel estimation method based on channel prediction and feedback are proposed separately,which are aiming to solve the problem of sparse channel estimation with uniformly sparse structure and time correlation and the problem of cluster channel estimation with cluster structure and time-varying information.Firstly,we study and analyze the characteristics of the underwater acoustic channel,including propagation loss,multipath propagation,Doppler effect and ambient noise.Then the common sparse underwater acoustic channel model and simulation analysis are presented.Based on the experimental data from Qiandao Lake,the actual channel is analyzed.The results show that the channel has obvious time-varying characteristics,the Doppler effect is an important factor affecting the temporal coherence,and the spatial coherence of the vertical channel depends on the structure of inter-array channels.Based on the experimental data from the South China Sea,the clustering characteristics of the sparse paths are analyzed,and the results show that the channel presents a typical sparse cluster structure,and the time variation among the cluster channels is independent,and the channel cluster structure with stronger energy is more stable with time variation.Secondly,without considering the prior distribution of sparse channels,the CS based low complexity sparse channel estimation methods have been studied.First an arbitrary grids-path identity(AG-PI)algorithm based on phase angle domain is proposed,to deal with the problem of constrained performance which is limited by the preset delay grid in Orthogonal Matching Pursuit(OMP)algorithm.Then an improved self search arbitrary grids-path identity algorithm(SSAG-PI)is proposed to solve the problem of performance fluctuation caused by improper angle domain search grid.In addition,under the system considering Inter-Carrier Interference(ICI),the OMP algorithm needs to consider both delay search and Doppler search,which have the problem of high computational complexity,an improved self search low complexity-OMP(SSLC-OMP)algorithm based on joint delay and Doppler search is proposed.Simulation and field experiment results show that the proposed algorithm can greatly reduce the computational complexity and achieve higher estimation accuracy under the same computational complexity.Then,under the consideration of sparse channel prior distribution and space-time dependent structure,the time correlation based sparse Bayesian learning channel estimation has been studied.The sparse Bayesian learning theory has been studied for sparse signal reconstruction by assuming parameterized prior distribution for the sparse solution and using the expected maximization algorithm to estimate the super parameters in an iterative way,the maximum posterior estimation of the sparse channel to be estimated could be obtained.Then the uniformly sparse structure and time-coherent properties of the solutions of continuous multi-block sparse channels are explored,and we propose the temporal multiple SBL(TMSBL)based channel estimator to jointly estimate the channels by taking advantage of the channel coherence between consecutive OFDM blocks.Simulation and field experiment results demonstrate the effectiveness of the SBL and TMSBL channel estimator algorithms in slowly time-varying UWA channel,which achieve better channel estimation performance and lower bit error rate compared with the existing CS-based methods,such as OMP and multiple OMP methods,especially the TMSBL estimator achieves the best performance in strong temporal correlated channels and maintains robustness in weak temporal correlated channels.Finally,under the consideration of the cluster structure and time-varying information of channels,the channel prediction and feedback based channel estimation methods have been studied.The channel prediction based joint channel estimation method is proposed.We define a channel offset parameters model by the clustering property of UWA channels,and reconstruct a virtual current received signal based on the prediction method.Then we combine the virtual received signal and the actual one into a joint estimation model,and utilize TMSBL method to jointly estimate the channel.In addition,due to the limited pilot resources in fast time-varying channels,the performance of channel estimation based on pilot assistance is contradictory to the system efficiency,we propose the decision feedback-joint SBL(DF-JSBL)algorithm for joint channel estimation and data detection,and use unknown data symbols as additional observations to improve performance.Simulation and field experiment results demonstrate the effectiveness of the channel prediction based joint channel estimation method in fast time-varying cluster UWA channels,which can effectively improve the robustness of the communication system,and the DF-JSBL algorithm can effectively reduce the bit error rate of the system and improve the bandwidth utilization and communication rate,and improve the performance of the communication system.