Research On Algorithms For Spatial Signal Adaptive Beamforming And Parameters Estimation

The adaptive beamforming and parameters estimation of spatial signal are two important research fields of array signal processing. They finds wide applications in many fields such as radar, sonar, seismic exploration, electronic surveillance, radio astronomy, etc. this dissertation aims at the problems of adaptive beamforming and estimation of signal parameters in different practical environments, provide several useful algorithms and verifies them by computer simulations. The main contents of this dissertation can be summarized as follows:1. Based on the data matrix, a novel algorithm termed bi-iteration algorithm (BIA) is proposed to implement the two-dimension (2-D) adaptive beamforming. The long weight vector of the Standard Capon beamforming (SCB) can be written as a kronecker product of two short weight vectors. The proposed BIA is used to deduce the two short weight vectors. Then the long weight vector can be reconstructed through the two short weight vectors by Kronecker product. Since the dimension of data matrix we use is much lower than the dimension of correlation matrix in the standard capon beamforming, we can obtain a better estimation of data matrix than correlation matrix in the case of less snapshots. Hence our BIA shows a better performance in the case of small snapshots. Moreover, the inverse of the high-dimensional correlation matrix is avoided by using the data matrix, so the computational complexity of BIA is remarkably reduced. A set of experiments is presented to compare the proposed BIA with the well known Capon beamforming in the performance of convergence speed, the beampattern, the output SINR versus input SNR, input snapshots and so on.2. A non-unitary joint diagonalization algorithm is proposed to estimate the 2-D frequencies embedded in additive Gaussian white noise. By exploiting the rotational invariance property of the 2-D data matrix, we establish four matrices which possess diagonal structures. Moreover, we expand the data matrix in temporal domain and derive a set of diagonal structure matrices. Consequently the 2-D frequencies are estimated by accomplishing the joint diagonalization of the group of data matrices. It is worth mentioning that the estimated 2-D frequencies can be paired automatically. The proposed algorithm eliminates the error propagation of the multistage decomposition algorithm because each iteration poses a typical least square problem with a unique closed solution. Hence the estimation accuracy is increased. The asymptotical convergence of the proposed algorithm is also proved.3. We develop a generalized version of non-unitary joint diagonalization algorithm, and then propose an algorithm to resolve the problems of direction of arrival and harmonic retrieval. The underlying rotational invariance among signal subspace induced by an array of sensors with a displacement invariance structure is exploited, a set of spatio-temporal correlation matrices possessing diagonal structures are introduced. To make the computational complexity lower and the convergence speed faster, we deduce a dimension-reduction matrix dealing with the set of correlation matrices. Using the structure information of the dimension-reduction correlation matrices, a novel cost function is proposed based on nonlinear least squares. Afterwards, a new approach termed tri-iterative algorithm (TIA) is derived for solving the cost function and estimating the direction of arrival of sources and the harmonic frequencies. Simulation results demonstrate the effectiveness of the proposed algorithm.4. An algorithm is proposed to track the azimuth and elevation angle in the case of arbitrary plane antenna array. Introducing instrumental variable can make this algorithm used in colored noise environment. It makes use of the rank one update model to construct two unconstrained cost functions, then the signal subspace can be obtained through the recursive least square solutions of the cost functions. Two proper approximations are made in the deduction to reduce the computational complexity. The tracked signal subspace are orthonormalized to obtain better performance. The computational complexity between the proposed algorithm and the well known EIV-PAST algorithm is detailly analyzed and compared. Simulations are conducted to show the tracking performance of the proposed algorithm and the EIV-PAST algorithm. Specially, we compare the error of signal subspace, the subspace angle and the error of orthonormality.