Abstract: Based on the ESPRIT (estimating signal parameter via rotational invariance techniques) method, an improved algorithm for beamforming was proposed by utilizing the subspace tracking algorithm in perturbation analyses to obtain ideal beamforming under the condition that the direction of expected signals is unknown or changed. In this algorithm, real-time tracking for the direction vector of the expected signals is realized due to the application of the subspace tracking algorithm, as a result, the pointing error of time-varying signals can be controlled. Numerical simulations show that beamforming obtained using the improved algorithm approaches to the beamforming under the ideal condition that the direction of the expected signals is known.