Wideband Beamforming Algorithm Of Underwater Acoustic Signal Based On CUDA

Ocean is rich in natural resources and plays an irreplaceable strategic role in Military field.The urgent need for marine exploration promotes the rapid development of underwater acoustic engineering.The beamforming has the capability of airspace filtering and can be applied to estimate target azimuth.It is not only the key technology of underwater acoustic system but also one of the most intensive computing parts of underwater acoustic signal processing.In recent years,the type of signal of underwater acoustic detection has changed from the traditional narrowband signal to the broadband signal which can carry more information.This change makes the calculation of the beamforming lager than before.However,the system has a high requirement for the speed of underwater signal processing.Therefore,it is of great practical value to apply the parallel processing technology to accelerate the speed of underwater acoustic system based on the graphics processor.In this thesis,a wideband beamforming method is proposed to obtain a constant beam width based on the study of narrowband signals and is applied and accelerated based on Computer Unified Device Architecture on GPU platform.Firstly,the theory of array signal processing and beamforming is studied.The mathematic model of the array is established.The received signal model and the array manifold are deduced.In terms of narrowband signal,beamforming,beam deflection control and beam optimization are discussed.The optimal weight vector of beamforming is solved under the constraint condition of the noiseless response of the signal.Simulation results verify the effectiveness of beam deflection control method and beam optimization method.The suppression effect on the direction of interference of the beamforming optimal weight vector is also proved.Then,the structure and characteristics of conventional broadband beamforming are discussed.The simulation results show that the different frequency components of the wideband signal will have different response after the same array so that the time domain distortion of the beamforming is produced.In order to solve this problem,a beam optimization method of controlling the sidelobe and the width of the main lobe is studied based on the Dolph-Chebyshev weight.At the same time,by analyzing the relationship between frequency and parameters of array,a beamforming method of constant beamwidth for octave signal is designed by nesting high and low subarray.The simulation results show that each of the two kinds of constant beamwidth beamforming methods has the same response at different frequencies in the direction of the main lobe region and will not cause distortion response in time domain.Finally,the parallel acceleration of beamforming algorithm is realized on GPU based on CUDA.After introducing the architecture and programming model of CUDA,the broadband beamforming algorithm is decomposed into basic calculation functions.The main functios with large computations which are Matrix multiplication and Fourier transform algorithm are implemented and optimized.In the end,the performance of airspace filtering and accuracy of target azimuth for beamforming are tested via field experiment.The parallel algorithm of beamforming is done and proved to have high performance.