Subspace-based parameter estimation for array signal processing
Signal parameter estimation from sensor array measurements or multiple channel time series observations is one of the most important problems in signal processing.^ The maximum likelihood (ML) principle provides a systematic way to solve this problem. Nevertheless, the existing ML estimators generally require an iterative search over a multidimensional parameter space to obtain parameter estimates and thus suffer from high computational cost. Usually there is no guarantee of global convergence.^ Parallel to the maximum likelihood techniques, the class of subspace-based algorithms has been proposed which includes MUSIC, Min-Norm, State-Space Realization and ESPRIT. These algorithms are non-iterative. Among them, ESPRIT has received much attention because of its features of robustness and computational efficiency. Unfortunately, like other non-iterative subspace-based algorithms mentioned above, ESPRIT is statistically inefficient. To improve upon ESPRIT, an algorithm called Multiple Invariance (MI) ESPRIT has been proposed. However, MI ESPRIT obtains the parameter estimates by minimizing a nonlinear criterion function, and an iterative multidimensional search is generally inevitable. Thus, the computational advantages of ESPRIT are lost for MI ESPRIT.^ This thesis conducts a study on parameter estimation for array signal processing when an ESPRIT array is used. A new non-iterative algorithm called Weighted Subspace Estimation (WSE) is derived. The performance of WSE is investigated analytically and experimentally. Results show that, the performance of WSE achieves or is very close to the Cramer-Rao bound over a wide range of signal-to-noise ratios (SNRs) and signal correlation. ^
Statistics|Engineering, Electronics and Electrical
"Subspace-based parameter estimation for array signal processing"
Dissertations and Master's Theses (Campus Access).