Estimation of source parameters by maximum likelihood and nonlinear regression

Statistical properties of certain parametric array processing methods are investigated. Asymptotic normality of Fourier-transformed sensor outputs for usual signal plus noise models is applied to define likelihood functions which have to be maximized for parameter estimation. In the first well known approach, the parameter structure is contained in the spectral density matrix of the outputs. The second likelihood function is conditional and results in a nonlinear regression problem. Since the likelihood equations are difficult to solve in general, properties of approximate solutions, for example Liggett's method, are of interest. Asymptotic distributions of the estimates and their approximations and results of some numerical experiments are discussed.