A Regularised Normalised Augmented Complex Least Mean Square algorithm

A Regularised Normalised Augmented Complex Least Mean Square (RNACLMS) algorithm is proposed for widely linear adaptive filtering in the complex domain C. Based on augmented complex statistics, the RNACLMS is shown to utilise complete second order information in C, thus being suitable to deal with both circular and noncircular complex signals. Furthermore, a gradient adaptive regularisation term makes the proposed algorithm exhibit enhanced robustness and convergence over the NACLMS algorithm. Simulations on circular and noncircular benchmark signals and on real-world noncircular wind signals support the analysis.