Improved Gerschgorin disk estimator for source enumeration with robustness against spatially non-uniform noise

In this study, a characteristic equation-based Gerschgorin disk estimator (CE-GDE) is proposed for source enumeration. In CE-GDE, the diagonal averages of the array output covariance matrix of a uniform linear array are used to form a new data matrix, whose rank equals the number of the incident signals. Then the signal number is estimated by detecting the rank of this matrix with the Gerschgorin disk estimator. Numerical examples show that CE-GDE surpasses existing methods in scenarios of both spatially uniform and non-uniform noise.