On the stability of recursive least squares and QR decomposition algorithm for adaptive filtering application

This paper investigates the stability and tractability of the recursive least squares (rls) and the qr decomposition for the recursive least squares (qrd-rls) techniques for adaptive filtering application using the mean weight error norm (mwen) instead of the conventional mean square error (mse). the analysis is based on the comparative performance with respect to numerical stability and tractability of incoming signals. the analysis depends on the effect of the different values of the filter coefficient  to these techniques. the simulation indicate that the rls procedure reduces tractability and is numerically unstable as the value of  increases while the qr decomposition for the recursive least squares showed numerical stability and tractability of incoming signals. the simulation also revealed that the rls technique produces high misadjument as the filter coefficient increases; the contrary is true for the qrdrls technique. in general,the rls technique converges faster than the qrd-rls technique. © 2015,pakistan academy of sciences.