Subspace separation method for ISAR imaging using the MUSIC algorithm

This paper presents a method for subspace dimension estimation which is critical for accurate ISAR image construction using the MUSIC method. The proposed method examines the distribution of correlation matrix eigenvalues after various amount of spatial smoothing to derive an ideal amount of spatial smoothing and the corresponding eigenvalue threshold. When used to separate the signal and noise subspaces, this eigenvalue threshold provides an accurate estimate of the number of scattering centers on the target. The proposed method is tested with simulated ISAR data and compared with a fixed-threshold method. Accuracy over SNR and number of scatterers is presented as well as example images.