Model Based DOA Estimation For Acoustic Vector Sensor Array

Compared with traditional pressure transducer,the acoustic vector sensor can additionally provide acoustic field's vector information for postposition signal processing system.Therefore it attracts much attention in underwater acoustic domain.Classical acoustic vector signal processing methods usually consider the ocean channel and background noise as ideal conditions;these approaches ignore the influence of complex and variant sea environment to processing performance.The model based processing combine physical model with signal processing,and utilize the sound field characteristic to improve the processing capability.At present,the model based processing is deployed to deal with sound pressure signals,and the application in acoustic vector signals is little.So this paper will quest for acoustic vector field model and acoustic vector signal model based spatial spectrum estimation algorithm building on conventional DOA methods.1.DOA method based on acoustic vector field polarization modelThe performance of subspace high resolution DOA algorithm is restricted by space random noise.Aim at this,polarization preconditioning is introduced to underwater acoustic signal processing,and a DOA model is designed based on polarization analysis.At first,the particle motion trajectory in shallow water condition is studied under two theories,one is ray theory and the other is normal mode.By the two theories,it can be proved that due to the influence of waveguide,the polarization phenomena exist in underwater sound field.At the foundation of above research,the polarization characteristics of signal and noise are analyzed.Then the time domain polarization filter is designed to suppress instantaneous impulse according to the difference between signal and noise's polarization characteristic.At last,construct data recovery model by polarization parameters to predict data where impulse occurs.The validity is verified by simulation and real data results.Time domain processing has poor noise suppression capability.In order to behave good performance in isotropy noise,DOA estimation must take full advantage of frequency information.Extend time domain polarization analysis to frequency domain,and design the frequency domain polarization filter(FDPF)function to reduce the influence of isotropy noise.Then substitute the original signals with weighted signals to construct the covariance matrix and estimate the bearing by conventional DOA methods.The final results of simulation and real data indicate that the polarization analysis based DOA does well when the power of noise is high.The two targets' identification performance of FDPF-DOA is better than conventional DOA.Furthermore,the above algorithm hardly needs any prior knowledge,and suits for the orientation and tracking of passive target.2.DOA methods based on acoustic vector signal modelThe super performance of high-order cumulant in array aperture enlargement and spatially coherent colored noise repression make it more fit for real sea environment.The only problem is that there is a great deal of redundancy in the four-order cumulant,so the computation burden rises.To overcome this problem,two redundancy removed methods are advised,one is direct erasure approach,and the other is response average method.Theory analysis shows that the dimension of covariance matrix is reduced by the above methods.Simulation research verifies that the second redundancy removed method behaves better than original four-order cumulant DOA.Although the statistic model can extend the aperture,it still exist some application boundedness.To obtain accurate DOA estimation,large size array should be adopted if experiment condition permits.Compared with the scalar sensor array,computation quantity of vector sensor array's DOA algorithms grows faster with the increase of element number.Due to the quaternion's specific operation rules,the problem of big computation burden and memory space can be resolved.At first the existing quaternion model in acoustic vector sensor is modified to conquer port starboard ambiguity.Then pressure-velocity quaternion model is proposed against the positive-negative ambiguity of quaternion ESPRIT in being.At last combine the pressure-velocity quaternion signals with non-spatial ESPRIT.By the theory analysis,it can be known that the advised quaternion methods accelerate the speed of processing and scarcely lose any information.The validity of the suggested approaches is demonstrated by the simulation results.