Simple and robust analytically derived variable step-size least mean squares algorithm for channel estimation

Abstract

A new variable step-size least mean squares (VSS-LMS) algorithm for the estimation of frequency-selective communications channels is herein presented. In contrast to previous works, in which the step-size adaptation is based on the instantaneous samples of the error signal, this algorithm is derived on the basis of analytical minimisation of the ensemble-averaged mean-square weight error. A very simple rule for step-size adaptation is obtained, using a small number of communication system parameters. This is another significant difference from other proposals, in which a large number of control parameters should be tuned for proper use. The algorithm here proposed is shown to be applicable to both time-varying and time-invariant scenarios. While the lack of a termination rule for step-size adaptation is a common characteristic of other schemes, the algorithm here presented adopts a criterion for stopping the step-size adaptation that assures optimal steady-state performance and leads to large computational savings. A simulation-based performance comparison with other VSS-LMS schemes is provided, including their application to maximum likelihood sequence estimation receivers using per survivor processing (MLSE/PSP). The results show that the algorithm proposed in this work has good performance characteristics and a very low computational cost, specially in the application to MLSE/PSP receivers. Besides, this algorithm is shown to be robust to changes in the signal-to-noise ratio (SNR).