Fast and Accurate Subspace Tracking Algorithm Based on Variable Forgetting Factor

In the non-stationary environment, compared with the traditional subspace estimation algorithm, the projection approximation subspace tracking (PAST) algorithm can achieve less computation and better subspace tracking performance. Aiming at the problem that the fixed forgetting factor in the PAST algorithm cannot balance subspace tracking accuracy and convergence speed at the same time, this paper presents a fast and accurate subspace tracking algorithm that employs a variable forgetting factor, which inversely correlates with subspace tracking error. The algorithm realizes the simultaneous optimization of accuracy of subspace estimation and speed of convergence. Simulation results validate its effectiveness.