Cumulant projection method for two-dimentional harmonic retrieval in colored non-gassian noise

The problem of concern here is the two-dimensional (2-D) harmonic retrieval in colored non-Gaussian noise. The 2-D cumulant projection theorem of complex linear non-Gaussian process is presented. The noise auto-correlation estimation is obtained based on this theorem by skillfully constructing a high-order cumulant of the complex noisy harmonic. Then a generalized eigen-value problem is solved to prewhiten the noise space. Finally 2-D MUSIC is employed to retrieve the 2-D harmonic frequencies. This method can effectively extract frequencies from colored non-Gaussian noise based on AR or ARMA model, even though the noise is symmetrically-distributed or there exists quadratic phase coupling. Theory and algorithm are illustrated by simulation examples