One-Dimensional Lattice Potential-Based Stochastic Resonance for Robust QRS Detection in Noisy Electrocardiogram

Detecting QRS waves with high sensitivity and precision in noisy electrocardiogram (ECG) recordings is crucial for cardiovascular disease monitoring. A main challenge is the in-band noise contamination that overlaps with the ECG signal spectrum, which is difficult to completely eliminate using traditional filtering methods. Inspired by solid-state physics, we propose a 1-D lattice potential (OLP)-based algorithm to enhance robustness against in-band noise. The algorithm first bandpass filtered the raw ECG recordings to mitigate out-of-band noise. Subsequently, the ECG signal undergoes nonlinear processing through the OLP with a variable damping coefficient. For the QRS complex, it experiences weak damping and is enhanced by bistable stochastic resonance (SR) effect, where noise energy transfers to the effective signal during transitions between two potential wells. In contrast, non-QRS segments are suppressed by applying strong damping. The enhanced ECG signal is then processed through a high-pass filter (HPF) and thresholding stage to locate QRS complexes. Testing on four different ECG databases shows that the ${F}1$ -score of our proposed OLP-based algorithm ranges from 98.87% to 99.99%, outperforming other state-of-the-art algorithms. Moreover, compared to traditional monostable SR-based algorithm, the OLP-based QRS detector induces a stronger SR effect, resulting in a greater signal-to-noise ratio (SNR) gain. Consequently, in the environment with large real-world artifact and noise level, the proposed OLP-based QRS detector can improve the noise margin for maintaining optimal performance by 133%–269% compared to traditional monostable SR-based QRS detector.