Estimation of Multipulse Series Driving of an All-Pole Transfer System : Application of Tapering Window in Singular Value Decomposition

Abstract

When a vibration is repeatedly driven, it is difficult to identify the multipulse driving series from the observed signal contaminated by additive noise; this is the case even when the transfer system is described by an all pole model and its characteristics are known because an unstable nonlinear optimization method, such as pole-zero modeling, is necessary. Thus, an inverse filter is ordinarily used. However, when the signal-to-noise ratio is low, even this method cannot estimate the multipulse series accurately. Therefore, we propose a stable method to estimate the multipulse driving series from the observed signal by using singular value decomposition and applying a tapering window. Finally, the principle is confirmed experimentally.