Windows and Volterra transfer function estimation

Abstract

The effects of conventional data windows on Volterra transfer function estimation are investigated. The input/output data for two known second-order systems are utilized to estimate the transfer functions, and the results are compared with true values. In addition, the use of window correction factors to offset the bias introduced into the higher-order moment spectra, by the fact that the data is attenuated at the beginning and end of a record, is investigated. In all cases, it is found that the rectangular window results in the smallest NMSE (normalized mean square error) for the estimated quadratic transfer functions.