Frequency Response Matrix Estimation From Partially Missing Data—for Periodic Inputs.

Multivariate nonparametric frequency response estimation is important in engineering. It allows one to get a quick insight into the dynamics of the system from input–output measurements. Sometimes, measurements can be missing due to faulty sensors or communication links. In this paper, we develop a method that estimates the frequency response matrix together with the missing samples from partially known data. In addition, the method can estimate the level of nonlinear (NL) contribution and the additive noise level of weakly NL systems when excited by a periodic excitation. Several special cases can be handled if the reference input is known, like samples missing at the inputs, noisy inputs, and identification in feedback. [ABSTRACT FROM PUBLISHER]