2.3. Signal Preprocessing

Noise is inevitable when collecting ECG data, and the main sources of noise are the power frequency interference and the baseline drift. To accurately extract the relevant information in the ECG signal, we need to remove the noise interference in the ECG signal before extraction. The power frequency interference is a noise problem caused by electromagnetic radiation in the surrounding environment during the operation of the equipment; its frequency is 50 Hz. The baseline drift refers to the instability of the working conditions of the instrument when measuring the ECG signal, leading to the baseline of the ECG signal being not a horizontal line but oscillating up and down. As shown in Figure 3a,b, two main sources of noise exist in the signal. The effects of these noises can be substantially removed by different digital filtering methods. Recently, the Blind Source Separation (BSS) techniques such as Independent Component Analysis (ICA) are wildly used for the artifact removal of the physiological signals [32,33,34], however, our signals are single channel signals while the BSS techniques requires multi-channels signals, thus, we choose the traditional methods to remove the noise. First, we applied a notch filter to remove the effect of the power frequency interference. As we can see from the ECG signal after removing the power frequency interference, shown in Figure 3c, and its spectrogram, shown in Figure 3d, the majority of the 50 Hz noise was removed. We utilized a zero-phase filter to correct the baseline drift; the effect is shown in Figure 3e,f.

ECG signals before and after filtering and their spectrograms.