Estimation of damped oscilations using adaline network

The estimation of the frequency and damping factor of a decaying exponential is a problem of prime importance. This paper presents an algorithm based on adaptive linear neural network (Adaline) for online estimation of the frequency and damping factor of an exponentially damped sinusoidal (EDS) signal. The Adaline structure is based on a linear recursive model of the real and imaginary parts of a complex EDS signal. Using this model, the inputs of Adaline network are real damped sinusoids. The unknown parameters of signal put in the weight coefficients of the Adaline network. Normalized least mean square algorithm is applied to train the weight coefficients. The proposed method is also verified for the parameters estimation of a complex EDS signal including complex DC offset component. Convergence analysis of the proposed method is also presented. Simulation results are presented to support the desirable performance of the algorithm in different situations.