Robust Kalman-based DOA estimator with variable fading memory
Hansu Kim, Koeng-Mo Sung, Jun-Seok Lim · The Journal of the Acoustical Society of America · 1999
In this paper, VFM (variable fading memory) is introduced to the robust Kalman-based DOA (direction of arrival) estimator [Han-Su Kim and Koeng-Mo Sung, Proc. 16th ICA/135th ASA 1, 1327–1328 (1998)] for DOA estimation of uncorrelated narrow-band plane waves under the impulsive noise environment. To introduce VFM properly into the Kalman filter, the proposed algorithm is derived from a modified least-squares criterion with VFM for DOA estimation. Since the derived VFM can be adaptively calculated with respect to the estimation error, the proposed algorithm has the effectiveness to the impulsive noise in stationary or nonstationary situations. The introduction of VFM enables the time-varying DOA to be estimated properly and prevents the estimated error caused by the impulsive outlier from propagating into future time instants, adaptively decreasing the importance of past observations. The proposed method can make the small perturbation to the parameters using the Huber function, named M-estimation, and adaptively reduce the estimation error propagation caused by past observations of the impulsive noise using VFF. Compared to the conventional Kalman algorithms, simulations show that the proposed algorithm is robuster than any other conventional Kalman algorithms to the impulsive noise.