A Robust Heartbeat Detector not Depending on ECG Sampling Rate
All ECG processing algorithms start with heart beat detection, making the reliability of this step crucial for the quality of the whole interpretation. This paper presents a regression-based QRS detector that is robust to singular outliers and high frequency noise present in real signals. Our algorithm consists of three steps: best fitted segments are determined in sliding windows of two different lengths, then a function of running angle between segments is calculated, and finally based on all such functions a probability of heart beat occurrence is derived and evaluated. The algorithm does not use signal filtering, and the moving window length and step do not rely on the sampling interval. Consequently, it allows for missing data and works equally well for a wide range of sampling frequencies.