On cross-correlogram estimators of the response function in continuous linear systems from discrete observations (Q4940242)

The authors consider a time-invariant causal continuous linear system with a response function \(H(t)\) with \(t\in R\). This means that the function \(H\) vanishes on the negative half-axis and the response of the system to an input signal \(x(t)\), \(t \in R\), has the form \[ y(t)=\int_{-\infty}^{t}H(t-s)x(s)ds. \] Statistical discrete-time sample cross-correlogram estimators of the response function \(H\) under perturbations by a separable stationary zero-mean Gaussian process are considered. Conditions for the asymptotic normality of finite-dimensional distributions are studied. The proof of the main result employs, in particular, the moment method. The basic technical tool is related to some inequalities for singular integrals which represent multiple convolutions with linked arguments.