A Low-Complexity R-peak Detection Based on Exponential Weight Mean-Variance for Wearable ECG Devices

R-peak detection is crucial for many applications that rely on electrocardiogram (ECG) analysis, such as heart rate variability, biometric identification, arrhythmia detection, and cardiac diagnosis. However, R-peak detection is a challenging task due to the variability and complexity of the ECG signal and the presence of noise and artifacts. Existing methods for R-peak detection may have limitations such as low accuracy, high complexity, or manual parameter tuning, especially for wearable devices. This paper proposes a novel method for R-peak detection that addresses both accuracy and low resource consumption. Our method has two stages: pre-processing and detection. The pre-processing stage is used to gain the value of the QRS-complex area and reduce the contribution of the non-QRS-complex area. The exponential weight mean-variance (EWMV) is used to meet these requirements. In the detection stage, a finite state machine (FSM) is employed to systematically analyze the processed signal and detect specific patterns associated with R-peaks, enabling accurate identification and extraction of these critical cardiac features. Our method does not require manual parameter tuning or additional hardware resources, making it suitable for ECG wearable devices. Several databases of Physionet (MITDB, NSRDB, STDB, Challenge 2014) are used to evaluate the method’s performance. By comparing the proposed method with previous algorithms for R-peak detection, the result shows that our method achieves high accuracy (99.66%) on MITDB and robustness under various conditions and outperforms existing methods regarding computational complexity and resource efficiency.