Blind source enumeration based on Gerschgorin Disk Estimator and virtual array extension

This paper addresses the problem of blind source enumeration in the colored and unbalanced noise environment, where the number of sources can be equal to or more than the number of observers. In the proposed algorithm, the Gerschgorin Disk Estimator (GDE) is used as the detection rule and fourth-order cumulant of the observed data is used to estimate Gerschgorins radii. Compared with the minimum description length (MDL) and Akaikes information criterion (AIC) methods, the proposed algorithm can work well even when the information of the noise model is incomplete. Besides, because of the virtual array extension capability of high-order statistics, the proposed algorithm is effective in both determined and underdetermined cases.