Study On Low-complexity Parameter Estimation Method For Distributed Sources

Distributed sources exist widely in radar, sonar and mobile communications fields. For example, in dense urban environments, some objects such as building, road and vehicle cause easily the multipath phenomena, and may lead to certain spatial spreading of the source energy. The traditional parameter estimation techniques based on point source model aren't applicable to distributed sources. Therefore, a number of investigators have proposed some models of distributed sources and corresponding parameter estimation methods. However, most of them involve multi-dimensional nonlinear optimization or multi-dimensional spectrum search. The computational complexities of them are too high to be suited to real-time processing and engineering realization. In addition, all of them are based on the assumption that distributed sources are static, and cann't be applied to track the nominal direction-of-arrivals (DOAs) of mobile distributed sources in real time.Aiming at the above problems, based on an overall analysis of the existing models and parameter estimation methods of distributed sources, this dissertation focuses on two problems: the low-complexity parameter estimation and fast nominal DOA tracking of distributed sources, and gains several novel research achievements.Our main works and innovations are summarized as follows:1. Based on an intensive study on coherently distributed (CD) source model, incoherently distributed (ID) source model and the existing space-time distributed source model, two new distributed source models: temporal-angular coherently distributed sources and temporal-angular incoherently distributed sources are proposed. The expressions of corresponding array received vector and array covariance matrix are derived.2. A low-complexity parameter estimation method for CD sources is proposed. Using the approximate rotational invariance property between two shifted subarrays in a uniform linear array (ULA) and the propagator method, the proposed method can realize the decoupled estimation of nominal DOA and angular spread. Avoiding two-dimensional (2-D) search and the eigendecomposition of the sample covariance matrix, our approach has a substantially reduced computational cost than the existing methods. Further, it can provide excellent performance even at low SNR.3. Two low-complexity methods for the 2-D DOA estimation of CD sources are proposed. Based on two approximate rotational invariance relations in the special double parallel ULAs, the first method estimates the nominal azimuths and elevations of CD sources by obtaining two corresponding rotational invariance matrices using the total least squares (TLS) method. Based on three approximate rotational invariance relations in the special treble parallel ULAs, the second method obtains the 2-D nominal DOA estimates of CD sources by estimating three corresponding rotational invariance matrices using the propagator method. Without spectrum-peak search, the above methods provide lower computational complexity than the traditional methods, and can be applied to the multisource scenario with different angular distributions.4. Two low-complexity parameter estimation methods for ID sources are proposed. The first method can realize the decoupled estimation of nominal DOA and angular spread by utilizing the phase information of the secondary diagonal elements of the sample covariance matrix and the first column vector of the noise-free covariance matrix. Without the eigendecomposition of the sample covariance matrix, the method can estimate two parameters of single ID source using only once one-dimensional (1-D) search. Therefore, it provides lower computational complexity. The second method exploits the approximate rotational invariance property between two closely spaced shifted ULAs. And it estimates the nominal DOAs of ID sources by using the weighted eigenvalues of the corresponding rotational invariance matrix which is estimated by propagator method. Without spectrum search and the eigendecomposition of the sample covariance matrix, our approach is computationally more attractive compared to the existing methods, and is robust to mismodeling.5. Two low-complexity parameter estimation methods for 2-D ID sources are proposed. Using L-shape array, the first method formulated two types of spatial cross-correlation functions (SCFs). Under small angular spread, the nominal azimuth and elevation are estimated by using the phase information of the SCFs. Without any search and matrix eigendecomposition, our approach has a substantially reduced computational. Moreover, it is also a robust estimator which doesn't depend on the angular distribution shape of the ID source. Based on two closely spaced parallel ULAs, the second method firstly obtains preliminary estimates of the nominal elevations. And then it transforms a four-dimensional parameter optimization problem into two 1-D parameter optimization problems and a 2-D parameter optimization problem with the help of preliminary values of the nominal elevations. The method reduces significantly the computational burden and can be used to deal with multiple ID sources.6. Three fast methods for the nominal DOA tracking of distributed sources are proposed. Among the methods, two methods based on the subspace updating idea are used to track, respectively, the nominal DOAs and 2-D nominal DOAs of CD sources. Avoiding the repeated eigendecomposition of the sample covariance matrix, they can track multiple sources with different angular distributions. The third method takes advantage of the maximum likelihood (ML) estimator and particle swarm algorithm, and is used to track the nominal DOAs of ID sources. The method does not require the knowledge of the effective dimension of the pseudosignal subspace of ID sources, and can exhibit excellent tracking performance even at low SNR.