Signal subspace fitting method based on the conjugate gradient

A new signal subspace fitting method based on the conjugate gradient is proposed. The new method builds the signal subspace estimator from the residual vectors of the conjugate gradient (CG) method. The initial vector of CG method is set to be depended on the incident angle of signals. Then subspace fitting method is used based on the estimated signal subspace and DOA is obtained from the best estimator. The location function of the proposed method only involves one dimensional search which means much less computation than the conventional signal subspace fitting algorithm. Simulations show that the proposed method perform less than classic weighted signal subspace fitting (WSSF) method, but offers improvements in estimation performance while compared with MUSIC and ESPRIT algorithms.