Higher order spectra and Volterra analysis of mechanical systems

This paper discusses how cross Higher Order Spectra (HOS) can be used to estimate elements of a Volterra model up to order three under the assumption that the input signal is Gaussian. It is demonstrated that to estimate higher order kernels the effects of lower order kernels should be removed and that this removal can be regarded as playing the same role as the normalisation of auto HOS. The technique is applied to synthetic and experimental data