Study Of High Resolution Multidimensional Signal Parameter Estimation And Its Fast Algorithm

The estimation of multidimensional signal parameter is an important aspect ofmodern signal processing, and it has been very rapid development in recent decades. Itsapplications involve array signal processing, non-destructive testing, biomedical andother aspects. This paper studies the multidimensional signal parameter estimation inwhite Gaussian noise environment. The main work of the thesis is as follows:1. Relevant background knowledge and some multidimensional signal parameterestimation algorithms and their simulations are given in chapter2of the thesis.2. In chapter3, a subspace based method for multidimensional frequency estimation ofmultiple damped sinusoids is proposed. The method combines the subspace andprincipal singular vectors utilization for modal analysis (PUMA). The idea of themethod is that construct2-D sliced matrices from multidimensional sinusoids, andthen the2-D sliced matrices are used to construct a set of covariance matrix to obtainthe SVD of each covariance matrix. The available PUMA method is employed toestimate the frequencies and the damped factors in each dimension. Moreover, wecan use existing matching algorithms to pair the multidimensional frequencyparameters. The proposed method is more computationally attractive than the twoexisting methods.3. A method based on the eigenvalues and eigenvectors of matrics is proposed formultidimensional frequency estimation of multiple sinusoids in chapter4. The coreidea of the algorithm is to estimate the two-dimensional parameter informationcontained in the new constructed matrix by using the eigenvalue and theeigenvectors’ corresponding relationship. And the two-dimensional parameter isautolly paired. The processing time of the algorithm is faster than IMDF and HOSVDalgorithm and performs better.4. In chapter5, we give a conclusion of this thesis and discuss the development of thisfield.