Efficient distortion removal in electrocardiograms by multichannel filter banks

Abstract

Electrocardiograms (ECG) are extensively used to monitor the electrical activities of the heart. It is a valuable tool, especially for the initial medical diagnosis of different heart diseases. However, because of the weak nature of signals, ECGs easily get adulterated with distortions such as noise and baseline fluctuations. Although Savitzky-Golay (SG) filters are often used to remove such distortions from ECGs, they have limited distortion filtering efficiencies. This is because in SG filters a polynomial curve fitting process is employed, and it is not possible to directly specify the distortion removal frequency bands. In contrast to SG filters, Multichannel Filter Banks (MFB) realized by stable digital impulse responses would have higher distortion filtering efficiencies due to the ability to remove only the undesirable distortions based on the frequency sub-band characteristics, while preserving the valuable features of the ECG signals. Thus, to remove the distortions in ECGs more efficiently, in this paper we have proposed and developed a new band-pass MFB filter, using a two-band filter bank as the building block. Using experimental techniques, the ECG signal distortion removal efficiencies for the proposed MFB and SG filters were estimated and compared by evaluating the output Signal to Noise Ratios (SNR) for different input signal SNR levels. Adulterated ECGs were cleaned using the MFB and SG filters. While having the ability to simultaneously remove different kinds of ECG signal distortions, significantly improved distortion removal efficiencies indicated by an average SNR increase of 3.2 dB in the filtered ECG signals were observed by the proposed MFB filter when compared with the SG filter.