Research On The Adaptive Identification And Equalization Technologies Of Sparse Underwater Acoustic Channel

Underwater acoustic communication system is an important tool of monitoring the ocean environment and obtaining the sea information. Underwater acoustic communication channel is an important obstacle of the development of the underwater acoustic communication. Some typical long-range underwater acoustic channel is time-varying sparse multi-path channel, this will cause inter-symbol interference. In addition, the non-stationary communication channel will cause the received signal suffering non-circular (NC) characteristic. Based on the sparse characteristics of underwater acoustic communication channel and the non-circular properties of the communication received signal, we study the identification and equalization technology of underwater acoustic channel in this paper.When deal with most of the sparse and multi-path channel, the traditional adaptive algorithms, such as LMS (Least Mean Square) algorithm, RLS (Recursive Least Square)algorithm and AP (Affine Projection) algorithm, will show some problem: (a) They using a fixed step-size to update the tap-weight, this will lead to a poor tracking ability. (b) They do not consider the sparse characteristic of the channel and this will show a performance of slow convergence speed and high noise, (c) They do not exploit the non-circular properties of the received signal. ; In order to overcome these problem, we study the adaptive system identification technology based on the LMS algorithm. First, we propose a l0-SH-LMS(Shrinkage linear LMS based on l0-norm) algorithm, it not only exploit the priori and posteriori error to calculate the various step-size to adapt to the time-varying channel, but also introduce a penalty that favors sparsity in the cost function to promote the applicability for sparse condition. Second, we proposed l0-SH-WL-LMS (Shrinkage and Widely linear LMS based on l0-norm) algorithm whose can make full use of the noncircular properties of the signals of interest to improve the tracking ability and estimation quality. Third, we show the convergence analysis of the l0-SH-WL-LMS algorithm and the result show that the proposed algorithms has the stable convergence condition. Last, channel identification and equalization algorithm simulation results also show that the proposed algorithm has a faster convergence rate and lower steady-state error than LMS algorithm and l0-LMS algorithm. The equalizer applying the proposed algorithms can also get lower bite error rate.Due to the high convergence rate of the AP algorithm than the LMS algorithm, we also take the same improvement measures for the AP algorithm in this article. First, we proposed the similar l0-SH-AP (Shrinkage Linear AP based on l0-norm) algorithm and l0-SH-WL-AP(Shrinkage and Widely Linear AP based on l0-norm) algorithm, which is suitable for the time-varying sparse multipath channel. Second,to reduce the computational complexity of AP-type algorithms, we also introduce DCD iterations to the proposed algorithms(l0-SL-DCD-AP and l0-SWL-DCD-AP) in this paper. At this point, the multiplication can be solved by bit flip, and the computational complexity.is greatly reduced. Third, we show similar analysis to prove the mean and mean square convergence ability of the improved algorithms. Last, the identification and equalization simulations of improved algorithm under the sparse channel show that the convergence rate and steady-state NMSE (Normalized Mean Square Error) performance is better than the AP algorithm and l0-AP algorithm. The low computational complexity versions of the proposed AP-type algorithms has similar performance with their counterparts. The equalization performance of the proposed algorithms is also improved.