Power-line Interference Elimination from ECG Signals Using Notch Filtration: A Quasi-real Time Version

Different types of notch filtration have been developed and used for power-line-interference (PLI) suppression in acquired and preprocessed ECG signals; most of them originate from the distant past. Generally, the traditional notch filters (NF) affect ECG frequency components around the rated PLI frequency. Recently Dotsinsky [1] published a successful off-line notch filtration providing almost total PLI elimination. Recordings taken from the AHA database are mixed by synthesized interference with frequency, which varies from 49 to 51 Hz within 20 s. They are subjected in a MATLAB environment to bidirectional band pass (BP) filtration with central frequency f0 = 50 Hz. Thus, the phase shift is compensated and the zero crossings of the extracted sinusoidal BP waves are identical to that of the PLI. In this study, the standard bi-directional BP is replaced with such one working in a quasi-real time thus allowing immediate tracking and interpretation of the ECG signals. A third 150 Hz PLI harmonic is also added as part of the disturbance to bring us closer to real ECG acquisition conditions. Further on, the variable sine wavelengths are measured between two consecutive zero crossings defined as the first positive BP sample following a series of negative ones. The inter-sample distances are equated to unity. The left and right lateral distances are calculated using similar triangles and aided to the inter-sample distances. The obtained fractal wave periods are converted into corresponding variable PL frequencies fi. Corrupted signals are then subjected to bidirectional notch filtration (NF) with a narrow stop-band using fi as a continuously updated central frequency. The carried out experiments show error of about 2 μV between the original and processed ECG signals with sampling rate Q = 5 kHz.