Response of accelerated fast RLS adaptive filters to linear chirps

The author considers the problem of tracking a linearly chirped sinusoid with additive noise with an accelerated fast recursive least squares (FRLS) adaptive filtering algorithm, which was proven to be as efficient as the least mean square (LMS) algorithm for the identification of time-varying filters. It is shown both theoretically and experimentally that the accelerated fast RLS algorithm has the same performance as the standard fast RLS (which is inferior to the LMS performance). Thus, the acceleration method does not improve the performance in the chirp tracking context; however, the choice of the forgetting factor is made more flexible