A New Robust Kalman Filter-Based Subspace Tracking Algorithm in an Impulsive Noise Environment

The conventional projection approximation subspace tracking (PAST) algorithm is based on the recursive least-squares algorithm, and its performance will degrade considerably when the subspace rapidly changes and the additive noise is impulsive. This brief proposes a new robust Kalman filter-based subspace tracking algorithm to overcome these two limitations of the PAST algorithm. It is based on a new extension of the adaptive Kalman filter with variable number of measurements (KFVNM) for tracking fast-varying subspace. Furthermore, M-estimation is incorporated into this KFVNM algorithm to combat the adverse effects of impulsive noise. Simulation results show that the robust KFVNM-based subspace tracking algorithm has a better performance than the PAST algorithm for tracking fast-varying subspace and in an impulsive noise environment.