Analysis of the direct and indirect a posteriori RLSL algorithm

The numerical properties of two implementations of the a posteriori recursi ve least squares algorithm, the direct and the indirect algorithms, are ana lysed in order to qualify their numerical trustworthiness, Although both im plementations include update recursions that have the potential for large r elative errors, the manner in which they are propagated into the update rec ursions tor the reflection coefficients is shown to differ. It is this diff erence that allows for the direct implementation to be more numerically tru stworthy with respect to the indirect implementation. Copyright (C) 2001 Jo hn Wiley & Sons, Ltd.