Spectrum estimation and adaptive denoising of eElectrocardiographic signals using Kalman filters

A method of time-varying parametric spectrum estimation from ECG sequences is presented. Model parameters are estimated recursively using a Kalman algorithm, which extracts the time-varying parameters and state variables of an ECG sequence, as well. We consider the noisy time- sequence generated by nonlinear auto regression, when the observations of the series contain measurement noise in addition to the signal. The spectrum estimates for each time instant then are obtained from the estimated model parameters. Proposed Kalman filter model turns to be adequate for either noise reduction or parameter estimation of processed sequence. Results thus obtained show better performance of Kalman-based filtration algorithm, in the sense of SNR and WDD distortion measurements, in comparison to conventional stationary spectrum estimation.