Adaptive unscented filtering technique and particle swarm optimization for estimation of non-stationary signal parameters

The paper presents an adaptive unscented Kalman filter (AUKF) for the estimation of non-stationary signal amplitude and frequency in the presence of significant noise and harmonics. The initial choice of the model and measurement error covariance matrices Q and R along with other UKF parameters is performed using a modified Particle Swarm Optimization (PSO) algorithm. Further to improve the tracking performance of the filter in the presence of noise the error covariance matrices Q and R are adapted iteratively. Various simulation results for time varying frequency of the signal reveal significant improvement in noise rejection and accuracy in obtaining the frequency and amplitude of the signal.