Real-Time Heart Rate Estimation Algorithm Based on Adaptive Spectrum Correction and Peak Localization

Photoplethysmography (PPG) is a widely used low-cost, noninvasive heart rate (HR) monitoring signal in smart wearable devices. However, the accuracy of HR measurement using PPG is often compromised due to motion artifact (MA). To address this issue, this article develops a novel HR estimation algorithm by employing an adaptive correction mechanism on the accuracy and location of the spectrum peaks. Initially, PPG and triaxis acceleration (ACC) signals are preprocessed to obtain the quality parameters and reconstructed by use of signal sparsity. Then, the dicrotic notch of the PPG signal and the energy of the ACC signal are selected as features to assess the interference degree caused by MA. Based on the interference level, an adaptive spectrum correction (ASC) model and the adjusting law are established to enable dynamic removal of MA and reduce motion-induced noise. Finally, an adaptive spectral peak locating algorithm is proposed to automatically regulate the localization logic according to the dynamic features of PPG and ACC signals. Comparative testing on the open dataset has validated the effectiveness of the proposed approach.