ONLINE MODAL IDENTIFICATION BY A NOVEL VARIABLESTEP SIZE ADAPTIVE BLIND SOURCES SEPARATION

This paper proposes a new variable step-size equivariant adaptive source separation viaindependence (EASI) algorithm for on-line blind modal identification of structures. The choice of stepsize that can fast initial convergence speed and low steady state error is one of the critical challengesin adaptive blind source separation. Through establishing a new nonlinear function between the stepsize and the separating indicator, the proposed algorithm adaptively updates its step size to match itsperformance to the dynamics of the input signals and unmixing matrix. Then the fast convergencespeed is obtained while the steady-state error is low. Simulation results of a benchmark structureverify the superior convergence and better steady-state performance of the proposed algorithm overconventional EASI with fixed step-size.

کلیدواژه ها:

Adaptive Blind Sources Separation ، Equivariant Adaptive Source Separation ViaIndependence ، On-Line Structural Identification ، Variable Step Size Adaptive Algorithm