Session PC6.5

Enhanced Kalman Filtering for Reduction of Powerline Interference of ECG Records Using a Delta Operator

LE Avendaño, JM Ferrero*, G Castellanos-Domínguez

Universitat Politècnica de València
Valencia, Spain

In this paper, an improved method of representation for sampled models based on delta transform is presented. Delta transform has shown to have superior numerical properties in digital signal processing. Particularly, it is discussed delta operator applied to Kalman filter in order to obtain enhanced power line parameters estimation for ECG records filtering. Compared with most known approaches, where it is considered that the parameters of powerline signal don’t change, the proposed modeling considers that these parameters are non stationary. The noisy ECG signal is modeled as the sum of two different signals: the first, related with 60 Hz interfering signal, which is modeled as a sinusoidal oscillator generated with delta operator, and the second, a first order auto regressive model of ECG signal. Parameters of this model are adaptively set by means of Kalman estimator, which minimizes prediction mean square error. Performance of the filter is analyzed with SNR, PRD and correlation index measures, and then compared with adaptive sinusoidal interference canceller, Wavelet denoising, estimation-subtraction and nonlinear adaptive filtering approaches. Tests are done on a database of Universidad Nacional de Colombia, containing records from normal and pathological ECGs from adult patients. Tests results have shown that the proposed methodology greatly improves ECG records for a reasonable computational cost increase. It was obtained an improvement between 20 dB and 30 dB for input ECG with SNR of -6 dB to 12 dB, also correlation index goes over 99.9% on almost any input SNR of ECG normal and abnormal signals.

(Abstract Control Number: 81)