Robust Heart Rate Monitoring From Wrist Pulse Signals Corrupted by Motion Artifacts Using Dual Microphones and Adaptive Filtering

This study addresses the degradation of radial arterial acoustic pulse signals during hand motion by proposing a novel heart rate (HR) detection algorithm, recursive minimum error entropy-variational mode extraction (RMEE-VME). The algorithm synchronously collects wrist acoustic signals using dual microphones, with one channel detecting the pulse waves and the other capturing motion noise. To suppress motion artifacts (MAs), an adaptive filtering algorithm based on the MEE criterion is employed. In addition, variational mode extraction and spectral peak tracking techniques are integrated to mitigate random interference and enhance HR estimation accuracy. Experimental results from 12 volunteers across five motion modes demonstrate the superior performance of RMEE-VME. The algorithm achieves an average absolute error, mean absolute error (MAE) of 2.16 beats per minute (bpm), outperforming recursive least square (RLS) filtering (3.51 bpm), and unfiltered signals (16.13 bpm). Each in high-interference scenarios, such as free hand movement, RMEE-VME maintains robustness, achieving an MAE of 2.60 bpm. Furthermore, it ensures stable real-time tracking with minimal HR fluctuations, achieving reliable performance despite the significant overlap between motion and HR frequencies. By integrating a dual-microphone acquisition system with spectral tracking, RMEE-VME enables high precision and continuous monitoring, demonstrating strong resistance to motion interference and high accuracy for next-generation wearable devices.