Matrix sign algorithm for sinusoidal frequency and DOA estimation problems

Fast algorithms based on the matrix sign function are developed to estimate the signal and noise subspaces of the sample correlation matrices. These subspaces are then utilized to develop high resolution methods such as MUSIC and ESPRIT for sinusoidal frequency and direction of arrival (DOA) problems. The main feature of these algorithms is that they generate subspaces that are parameterized by the signal-to-noise ratio (SNR). Significant computational saving will be obtained due to the fast convergence of these higher order iterations and to the fact that subspaces rather than individual eigenvectors are actually computed. Simulations showing the performance of these methods were also presented