Instrumental variable subspace tracking with applications to sensor array processing and frequency estimation

Recursive methods for subspace tracking with applications to 'on-line' direction of arrival estimation, have lately drawn considerable interest. Instrumental variable (IV) generalizations of the projection approximation subspace tracking (PAST) algorithm are proposed. The IV-approach is motivated by the fact that PAST delivers biased estimates when the noise vectors are not spatially white. The resulting basic IV-algorithm has a computational complexity of 3mn+O(n/sup 2/) complex multiplications, where m is the dimension of the measurement vector and n is the subspace dimension. The performance of the proposed algorithms in tracking sinusoids in colored noise is illustrated by computer simulations.