Multichannel ARMA Parameter Estimation

Using a perturbation matrix, we introduced a hyperplane used to define a generalized null-spectrum, based on both the signal and noise subspaces, while the MUSIC and Min-Norm null-spectra are defined based only on the noise subspace. With the generalized nullspectrum, we derived the upper and lower bounds of a class of the generalized null-spectrum, called the maximum and minimum null-spectra, respectively. From a brief analysis of a class of the generalized null-spectrum and a geometrical interpretation, we expected that the resolution capability of the maximum null-spectrum would be superior to that of other null-spectra. By computer simulation it is shown that the maximum null-spectrum has higher resolution capability than other null-spectra. In addition, it is seen that the asymptotic distributions of the estimates of the direction of arrival by using the maximum and multiple signal classification null-spectra are the same.