Detection of hidden periodicities in models with discrete time and long range dependent random noise

Abstract

Trigonometric regression models take a special place among various models of nonlinear regression analysis and signal processing theory. The problem of estimating the parameters of such models is called the problem of detecting hidden periodicities, and it has many applications in natural and technical sciences. The paper is devoted to the study of the problem of detecting hidden periodicities in the case when we observe only one harmonic oscillation with discrete time, where random noise is a local functional of Gaussian random sequence with singular spectrum. In particular, the random sequence in the model can be strongly dependent. For estimation of unknown parameters the periodogram estimator is chosen. Sufficient conditions of the consistency of the amplitude and angular frequency periodogram estimator of the model described above are obtained in the paper. The proof of Lemmas 1 and 2 gave an important asymptotic properties of the random noise functional related to the periodogram estimator which necessary for the proof of the main results. Series expansion of random noise in terms of Hermite polynomials and the Diagram formula are main tools that were used to obtain this lemmas.