No label defined (Q4270741)

One model often used in signal processing, system identification and time series analysis is the model of several real sinusoids in noise. It's main parameters are frequencies. The next popular model is a noisy sum of several damped exponentials. The main parameters of such a model are the damping factors. A combination of these two models is a noisy sum of damped sinusoids. It's main parameters are the damping factors and frequencies. In this work a model which generalizes all three models mentioned above is considered. This is the so-called multiple exponential-sinusoidal model. It consists of a sum of several multiple damped exponentials and multiple damped sinusoids. The paper deals with the problem of estimation of parameters of multiple exponential-sinusoidal models. It is shown that this problem reduces to a separable nonlinear least squares problem. Problems of implementation of the Levenberg iterative procedure for estimation of ``nonlinear'' parameters of the model under study are investigated.