A Fast Hybrid Adaptive Filter For Parameter Estimation Of Non Stationary Sinusoid Under Noise

摘要

This paper presents a modified combined approach using Taylor series expansion and Extended Kalman filter for accurate estimation of parameters and harmonic components of a time-varying signal embedded in noise with low signal-to-noise ratio. The signal is modelled using a dynamic model with time varying parameters. The power signal with a changing envelope has been expressed using a second order Taylor expansion. EKF algorithm is used for computing the parameters of such signal model. The algorithm using a linear model approach is presented which also reduces the computational complexity. Moreover this approach is also immune to noise, harmonic contaminations and also shows better convergence properties.