Variable Step-Size Transform Domain ILMS and DLMS algorithms with system identification over adaptive networks

Abstract

This paper presents a powerful performance and convergence speed of Variable Step-Size Transform Domain Incremental/Diffusion Least Mean Square (VSS-TD-I/D-LMS). It modifies and extends several already existing algorithms of VSS-LMS and VSS-TD-LMS to wireless sensor adaptive networks. The effect of transform domain along with power normalization plays a rule in reduce eigenvalue spread of input autocorrelation and whitening the highly correlated process. In ILMS, each node sensor is allowed to share its estimate with a direct neighbor while in DLMS each node update its estimate a long with a group of neighbors. Simulation results are shown that the performance improvement of cooperative fashion has substantial and favorable convergence speed. Simulation results are shown the performance improvement of cooperative fashion in convergence speed.