Real-Time Baseline Wander Removal in ECG Signal Based on Weighted Local Linear Regression Smoothing

Removing the baseline wander (BW) is vital in electrocardiogram (ECG) preprocessing steps, since it can severely influence the diagnostic results, especially in computer based diagnoses. This paper presents a method based on weighted local regression smoothing to correct BW in real time. Each signal data sample within a certain window is weighted. The weight of each sample is determined by the distance between the sample and the to-be-predicted sample. Then the regression is adopted by performing linear least-squares and a polynomial model to estimate BW. The ECG signal free from BW is obtained by subtracting the BW from the original ECG signal. The experiment results demonstrate that this method can effectively remove BW in ECG signal in real time and with minimum distortion of ECG waveform.