Mean Weight Behavior for the Generalized Subband Decomposition LMS Algorithm

This paper presents an improved stochastic model for the generalized subband decomposition least-mean-square (GSD-LMS) algorithm. This algorithm is used as an alternative to the standard LMS aiming to improve the convergence speed under correlated input data. An analytical model for the first moment of the adaptive filter weights is derived considering just the independence between weight and input data vectors. Numerical simulation results confirm the accuracy of the proposed model, outperforming other models presented previously in the literature.