Mean-square error analysis of the LMS-RVSS adaptive algorithm

This work presents the mean square error analysis of the Robust Variable Step Size (RVSS), which has been recently proposed. An analytical model is derived for the excess mean square error for white Gaussian input signals and slow adaptation. As a side benefit of the analysis, a new theoretical model is also obtained for the classical VSS algorithm. Monte Carlo simulations verify the accuracy of the analytical model. A practical example with real nonstation-ary biomedical signals illustrates the applicability of the algorithm.