A Nonquadratic Algorithm Based on the Extended Recursive Least-Squares Algorithm

In adaptive filters, several recursive algorithms have been used to track state-space model vectors in nonstationary environments. So far, kernel recursive algorithms showed the best results in this regard. With this letter, we aim to propose an algorithm based on a nonlinear function of the error, motivated by the extended recursive least-squares algorithm. Simulations were performed on the problem of tracking a nonlinear Rayleigh fading multipath channel and on a system identification. The results showed that the proposed algorithm can overcome the extended kernel version ones.