An Adaptive Linear Neural Network for Identification of Oscillatory Damped Signals

This paper presents an algorithm based on adaptive linear neural network for online estimation of damping factor and frequency of a complex exponentiallydamped sinusoidal signal. The unknown parameters of signal put in the single weight of a neural network. Normalized least mean square algorithm in complex form isapplied to train this single weight. A variable step size is proposed to enhance the accuracy and convergence speed ofthe proposed method. Convergence analysis of the proposed method is presented. Simulations results confirm the analytical derivations and desirable performance of the proposed method

Keywords:

adaptive linear neural network (Adaline) , complex exponentially damped sinusoidal (EDS) signal , complex least mean square algorithm , normalized least mean square , variable step size