Study On The High Resolution DOA Estimation Techniques Of Underwater Targets

High resolution array processing is a new quickly developed branch in signal processing during the recent two decades. It has very important applications in the fields of radar, sonar, communication, medical engineering, earthquake and so on. With further investigation on the high resolution technique and higher and higher application requirements in underwater acoustic signal processing, it is very important to study the new theory and methods of high resolution array processing and solve the complex problems of underwater acoustic signal processing to improve the performance of underwater equipments.Aiming at the application of underwater equipments, the new theory and method of high resolution DOA estimation and its experiments are studied in this thesis. A number of important results are obtained to provide the advanced methods and effective technique for the new underwater equipments. The main creative contribution in this thesis is as follows:1. The Cramer-Rao bound (CRB) of narrow band DOA estimation and the statistical performance of maximum likelihood DOA estimator (MLE) are studied. The results provide a theoretical base for the theoretical study and performance analysis of high resolution algorithms. The performance of high resolution subspace algorithms is compared with the CRB and MLE. And the run time in DSP hardware system for subspace algorithms is measured. The results show that the subspace algorithms possess high resolution and precision for DOA estimation and good prospect in real-time processing of underwater equipments.2. The theory and methods of beam space high resolution DOA estimation are studied. The performance of beam space high resolution algorithm is investigated based on the analysis of the gain of multibeam array system, which reveals the reason why beam space algorithm can obtain the improved performance. The results illuminate the relationship between the performance of beam space algorithm and the key factors such as beam number, beam overlap and weighting form. The criterion to determine the beam number and beam overlap is also given. The statistical characteristics of the mean square error and resolution probability of the beam space DOA estimation algorithms are obtained by computer simulation.3. The discrete frequency domain model of wide band array signal is established. The sample principle of wide band array signals in frequency domain is determined by applying the frequency domain sampling theorem. The CRB of wide band DOA estimation is proposed. The mathematical expression for the CRB of wide band DOA estimation is also presented. The CRBs of wide band DOA estimation for the different directions are calculated to provide an important benchmark to study performance of wide band high resolution algorithms of DOA estimation.4. The wide band coherent signal-subspace (CSS) method and its focusing matrixes are studied. Based on the statistical performance investigation of CSS using subspace disturbing analysis method, the relationship between the DOA estimation error and the signal to noise ratio (SNR) is presented, which reveals the statistical characteristics of the CSS. The research results show that wide band CSS method possesses good performancedue to resolving coherent sources and anti reverberation ability.5. By analysis of the covariance matrix eigendecomposition structure of wide-band array signals, the noise-subspace invariability of wide-band signals with the frequency is proved. The wide-band signal-subspace spatial-spectrum estimation based on frequency domain modeling (SSEFD) is proposed by using the noise-subspace invariability. The SSEFD shows superior performance over the CSS in DOA estimation precision.6. Wide band direct processing methods bring with them the problem of signal-subspace dimension expanding, which makes it difficult to correctly partition the signal-subspace and noise-subspace for high resolution DOA estimation. Three modified methods are proposed to overcome the signal-subspace dimension expanding. These methods are respectively named the DOA estimator based on signal-subspace dimension estimation, weighted noise-subspace DOA estimator and extended MUSIC method. Simulation results show efficiency of the modified methods to overcome the effect of signal-subspace dimension expanding and to improve the high resolution and estimation accuracy.7. The frequency-dependent modeling (FDM) algorithm for wide band DOA estimation increase the dimension of covariance matrix of array signal to represent noise subspace effectively and to improve the performance. However FDM bring it with very heavy computational burden. The coherent signal-subspace focusing to frequency-dependent modeling (CSSFDM) is proposed by introducing the focusing into FDM. Three focusing matrixes of CSSFDM are also constructed. The CSSFDM not only retain the good performance of FDM, but also reduce its computational complexity obviously.8. Based on the constant beam-width beamforming, wide band beam space high resolution algorithms are discussed. The theory and methods of wide band beam space high resolution estimation is studied. Results show that the wide band beam space MUSIC algorithm has better performance than the CSS in resolution and computation complex. Based on the analysts of noise property in constant beam-width beam domain, a weighted noise subspace DOA estimator in constant beam-width beam space is proposed. This method overcomes the effect of the beam space color noise and improves the performance of wide band beam space MUSIC algorithm.9. A high resolution DOA estimation experimental system in water tank is constructed. A number of high resolution methods, including narrow band, wide band, array space and beam space, are tested by using the experimental system. Experimental results show that the high resolution methods presented above successfully break through Rayleigh limitation (one beam width) and the resolution of DOA estimation is less than the half of beam width with the good robustness. The results also show the correctness and efficiency of these high resolution methods, as well as its promising engineering applications.