Frequency-domain quaternion-valued adaptive filtering and its application to wind profile prediction

Abstract

A multi-channel frequency-domain quaternion-valued adaptive filtering structure is proposed based on the quaternion-valued discrete Fourier transform (DFT). Similar to the traditional frequency-domain methods, the proposed method can improve the convergence rate of the employed adaptive algorithms, leading to much better tracking capability for applications in dynamic environments. Simulations based on an augmented quaternion-valued least mean square algorithm are performed for wind profile prediction using recorded data from multiple measurement sites, clearly demonstrating the potential of the proposed method in this specific area.