2-DIMENSIONAL MODAL-ANALYSIS BASED ON MAXIMUM-LIKELIHOOD

We present a method for estimating the parameters of 2-D damped harmonic (modal) signals in additive Gaussian noise. This method applies in the cases of single or multiexperiment data and incoherent or coherent signals, Both synthesis and analysis models for the data are developed. Central to these models are the signal and orthogonal subspaces. Under the assumption of distinct modes in one of the two dimensions, there exists an invertible function between the parameters of these two subspaces. This function can be used to express the optimization problem given by maximum likelihood in terms of the orthogonal subspace. The resulting problem can be solved by any of several methods, including total least squares and iterative quadratic maximum likelihood (IQML).