Abstract
This paper considers the development of fast widely-linear (FWL) recursive Least-squares (RLS) algorithm well suited for processing non-circular signals. The proposed algorithm makes use of covariance and modified covariance matrices which take full advantage of second order statistics of non-circular data. Further, the proposed algorithm is based on the fast QR-decomposition recursive least-squares (QRD-RLS) algorithm. Therefore, its computational complexity is of O(N) as compared to O(N-2) of conventional WL-RLS and is numerically more stable in finite precision environment. Simulation results have been presented to test the proposed FWL-QRD-RLS algorithm in two adaptive filtering scenarios: system identification and uniform array beamformer.