Research On Orthogonal Matching Pursuit Based High-Performance Sparse Channel Estimation In Underwater Acoustic OFDM Communication

Due to its high communication rate and strong ability to resist frequency selective fading,orthogonal frequency division multiplexing(OFDM)has become an effective solution to alleviate the disadvantages of underwater acoustic(UWA)communication.However,the multipath and Doppler effect of UWA channel often limit the advantages of OFDM.Doppler effect can be estimated and compensated with high accuracy by means of the autocorrelation of cyclic prefix of OFDM signal and digital signal processing(DSP)technologies.Based on this,this paper focuses on the sparse channel estimation caused by multipath effect in UWA OFDM communication system.In this paper,the orthogonal matching pursuit(OMP)sparse reconstruction model is utilized as the channel estimation framework.Aiming to solve the problems of the large amount of inner product calculation caused by high oversampling factor,the basis mismatch caused by parameter discretization and the strong interference caused by small multipath delay interval,the improved algorithms based on the efficient inner product searching strategy,the interpolation of the convex function and the effect of path fitting are studied,which provide the channel estimation schemes with low complexity and high reliability for UWA OFDM system.Firstly,there is a trade-off between the inner product calculation and the estimation accuracy in compressed sensing algorithms.The long delay spread of UWA channel also leads to the redundancy of computation.Based on OMP sparse reconstruction framework,this paper proposes two low-complexity OMP channel estimation models combined with the efficient DSP algorithms and searching strategy.This paper proposes a channel estimation scheme based on two-stage scheme facilitated by Goertzel’s algorithm,which reduces the inner product computational complexity in typical multipath sparse channels.This searching scheme can concentrate the cost of inner product calculation locally,and make use of Goertzel’s algorithm to replace the traditional inverse fast Fourier transform(IFFT).To reduce the computational complexity of inner product calculation in clustered channel,another OMP channel estimation model based on the cluster switch and Chirp-Z Transform(CZT)is proposed in this paper.In this model,CZT combined with the compressibility of clustered sparse channel is utilized to calculate the inner product function and search the peak of the function cluster by cluster.A threshold is set to realize the cluster switch to avoid overfitting within the cluster,which saves the computing resources greatly.Simulation and complexity analysis show that the two proposed algorithms can reduce the computational complexity significantly while maintaining the stable estimation performance of OMP.Then,the accurate and efficient path delay estimation model within weak inter-path interference is studied,in order to solve the performance degradation caused by off-grid path delay when small oversampling factor is adopted.Based on the proposed low complexity optimization model,the mathematical mechanism hidden in OFDM signal and OMP model is fully explored,and the traditional peak searching scheme is transformed into the extremum solving problem of the convex function.Three path delay estimation algorithms based on interpolation are proposed to improve the reconstruction accuracy of channel frequency response by improving the accuracy of each path delay.Among them,the unbiased closed-form estimate of path delay based on trigonometric function breaks the grid limitation of OMP,and creates a new direction of delay estimation for UWA path delay estimation.Simulation and complexity analysis show the advantages of the proposed algorithms in both estimation accuracy and computational complexity.Finally,based on the proposed low complexity OMP model and the interpolation idea,this paper puts forward the idea of path compensation to improve the channel estimation in the conditions of off-grid delay and strong inter-path interference.The compensation idea brings obvious performance gain to OMP based algorithms.In the proposed scheme,the expression of the coefficient of determination is constructed by the compensation distance,and the monotonicity trend of the coefficient of determination is proved.According to this trend,two improved algorithms based on the linear fitting of two atoms are proposed to alleviate the strong inter-path interference.The advantage of the proposed idea of path compensation and the improved algorithms are validated in the simulations and complexity analysis.According to the low complexity strategy,the interpolation idea,the path compensation and the double-atom fitting idea proposed in this paper,the performance evaluation based on the experiment received signal and experiment channel information is carried out.The comparison of the raw bit error rate(BER)corresponding to received OFDM signal collected by vertical line array shows that the interpolation idea,the idea of path compensation and the double-atom fitting idea have advantages in improving the communication performance of the system.By analyzing the decoded BER of various algorithms based on different pilot patterns in the verified channel,it is proved that the decoding performance of the proposed double-atom fitting algorithm has an edge over most classical sparse parameter estimation models in the strong inter-path interference channel.The adaptability to pilot patterns of the proposed methods is also verified.