Some Researches On Array Signal Processing

Array signal processing often used in many diverse fields of science and engineering such as radar, sonar, communications, radio astronomy, medical diagnosis, seismic exploration, military electromagnetic countermeasures (ECM) is an important research area of signal processing. Though the fundamental theories and basic methods have been presented in recent three decades development for array signal processing research, there are still many issues to be solved in the practical applications. This dissertation offers an in-depth study for some of these issues, i.e., estimation of the number of sources at low signal-to-noise ratio (SNR), joint estimation for multi-parameter, direction-of-arrival (DOA) estimation for coherent signals and DOA estimation for wideband sources, and then proposes a series of effective methods to deal with these issues mentioned above. The theoretical analysis is confirmed through numerical examples. Therefore, the main contents are as follows:The conventional eigenvalue-based techniques for estimating the number of the sources are studied, and the characteristics of the array correlation matrix are also analyzed at low signal-to-noise ratio (SNR). It has been noticed that the variance of noise may corrupt correlation matrix eigenvalues, but only have slight effects on eigenvectors, especially at low SNR, which makes the eigenvalue-based techniques lose the performance advantages like high SNR case. Therefore, this dissertation proposes two eigenvector-based methods according to the practical low SNR circumstances such as electron reconnaissance. In addition, the multivariate analysis methods are introduced to detect the signal at low SNR. Finally, simulation results that the proposed methods are superior to the traditional eigenvalue-based methods at low SNR without compromise of their high SNR case.We study the DOA estimation problems of joint elevation and azimuth and pay attention to analyse the L-shaped array widely applied in practical situations, then some methods are proposed to deal with these problems according to the different types of signals by taking good advantage of the correlation matrix. Firstly, we present a fast and efficient method of which the statistical analysis is studied with respect to the noncoherent signals, and the asymptotic mean-squared-error (MSE) expression of the estimation error is derived. Secondly, two efficient methods are suggested without pairing match problems when the signals are uncorrelated. Finally, we propose a new 2-D DOAs estimation method for coherent narrowband signals. Consequently, the main contributions of this chapter aim at putting forward some new methods to cope with the practical 2-D estimation problems under the different signal conditions using L-shape arrays.DOA estimation for coherent signals is studied, and a fast and efficient DOA estimation algorithm with known waveforms is proposed based on the traditional algorithms. The proposed method obtains the required subspace by carrying out the linear operation for the array data without eigendecomposition and estimates the DOAs by exploiting the shift invariance property of the array geometry without searching over the parameter space or finding polynomial roots, of which the computational complexity is high. Simulation results show that the proposed method is computationally much more efficient than the previous methods and only slightly less accurate. Therefore, our method can be easily implemented in practical situations.We study and analyse the problem to improve the capability for detecting the signals and the performance of parameter estimation by making use of the temporal and spatial information of signals. At first, a method for joint angle-frequency estimation is proposed. A pseudocovariance matrix without the effect of additive noise is constructed by take advantage of "sufficient" temporal correlations, and then the parameters are estimated with automatic pairing by using the matrix. Next, an efficient array expanded method for DOA estimation is also presented with respect to the minimum redundancy arrays. In addition, we propose a method for joint elevation and azimuth angles estimation based on the commonly used L-shape arrays. Finally, numerical simulations show that the signal temporal information can be utilized to improve both the number of signals that can be detected and the estimation accuracy.The DOA estimation techniques for wideband sources ubiquity in real-life are studied and analysed and a novel scheme is proposed according to some cases. Like the traditional wideband methods, this algorithm first decomposes wideband sources into narrowband bins. Then, at each frequency bin, the cross-correlation between sensors is estimated to construct a Toeplitz matrix to estimate the DOAs for wideband sources, where the effect of additive noise is alleviated. Simulation results show that the proposed technique can deal with the scenarios where the number of sensors is smaller than that of the sources.