An enhanced optimal windowed RLS algorithm for fading multipath channel estimation

Estimation of a fading channel with high accuracy is a key factor for successful signal detection in many mobile, indoor and personal radio communication systems. The optimal windowed recursive least squares (OW-RLS) algorithm developed previously (1993) results in a significant improvement over an exponential windowed recursive least squares algorithm when it is applied to tracking a fading channel. The improvement is confirmed by simulation on a single path fading channel. However, no improvement in performance has been obtained in the simulation of multipath frequency selective channels. This paper shows that the undesirable result is caused by an additional error which is introduced by an incomplete decoupling of cross-correlation terms. The authors design a feedback decoupling method which can cancel the bias almost completely. Simulations show that the theoretical improvement of an OW-RLS algorithm on a fading multipath channel can be completely recovered. With this modification, an enhanced OW-RLS algorithm is described.<
>