A Novel Adaptive Spectrum Noise Cancellation Approach for Enhancing Heartbeat Rate Monitoring in a Wearable Device

Yang D., Cheng Y., Zhu J., Xue D., Abt G.A., Ye H., Peng Y.

This paper introduces a novel method called Adaptive Spectrum Noise Cancellation (ASNC) designed to remove motion artifacts in Photoplethysmography (PPG) signals obtained from an optical biosensor. These artifacts often occur when the user is in motion, particularly when the motion frequency is similar to the target heartbeat rate. ASNC utilizes onboard accelerometer and gyroscope sensors to detect and adaptively remove these artifacts, thus enabling accurate heartbeat rate measurement during movement.

The ASNC algorithm employs discrete cosine transform, a widely used spectrum analysis approach in medical digital signal processing, to perform frequency domain analysis. Results from ASNC were compared to traditional algorithms, namely adaptive threshold peak detection and adaptive noise cancellation. The mean absolute error and mean relative error of heartbeat rate calculated by ASNC were found to be 0.33 (0.57) beats∙min^-1 and 0.65%, respectively. In comparison, the adaptive threshold peak detection algorithm yielded 2.29 (2.21) beats∙min^-1 and 8.38% error, while the adaptive noise cancellation algorithm resulted in 1.70 (1.50) beats∙min^-1 and 2.02% error.

While all algorithms performed satisfactorily with simulated and clean PPG data, ASNC demonstrated superior accuracy in scenarios with increased motion artifacts, particularly when motion frequency closely aligned with the heartbeat rate.

View this research