The Use of Wrist EMG Increases the PPG Heart Rate Accuracy in Smartwatches

Severi Friman, Antti Vehkaoja, and Jose Maria Perez-Macias

Published in IEEE Sensors Journal, vol. 22, no. 24, pp. 24197–24204, December 15, 2022. DOI: 10.1109/JSEN.2022.3219297 | IEEE Xplore: 9942941

Overview

Smartwatches and fitness wristbands predominantly rely on optical photoplethysmography (PPG) to non-invasively monitor heart rate (HR). However, during physical activities such as running, cycling, or weightlifting, voluntary muscle contractions in the wrist and forearm deform blood vessels and shift the optical sensor interface, inducing severe motion artifacts (MAs).

In this study, we demonstrate that incorporating surface electromyography (sEMG) measured directly at the wrist provides a direct reference for muscular motion noise. By fusing wrist sEMG with optical PPG signals, the proposed framework substantially improves heart rate estimation accuracy compared to traditional single-modality or accelerometer-only approaches.

The Challenge: Motion Artifacts in Wrist Photoplethysmography (PPG)

While photoplethysmography functions well when the user is sedentary or asleep, hand movements and grip forces cause localized muscular activation in the flexor and extensor muscle groups. These contractions:

• Modulate optical transmission through subcutaneous tissue, producing false rhythmic peaks that mimic cardiac pulses.
• Frequently overlap with the true heart rate frequency in the spectral domain, deceiving classical peak-detection and adaptive filtering algorithms.
• Cannot always be captured accurately by triaxial accelerometers alone, because isometric muscle contractions (e.g. gripping handlebars or weights) generate high internal muscle tension with minimal limb acceleration.

The Proposed Method: Multi-Modal Wrist EMG & PPG Fusion

We investigated integrating surface electromyography electrodes into the wristband configuration alongside standard green-light optical PPG sensors:

• EMG Noise Referencing: Wrist sEMG captures electrical potentials generated by the flexor carpi radialis and extensor digitorum muscles during hand and finger gestures.
• Signal Demodulation & Noise Removal: The EMG signal serves as a physiological noise reference, enabling adaptive filters to isolate and cancel contraction-induced disturbances from the raw optical signal.
• Clean Pulse Extraction: The filtered PPG signal preserves the true cardiac pulse waveform, yielding reliable inter-beat intervals and heart rate metrics.

Key Findings & Significance

Experimental evaluations across dynamic exercise protocols verified that:

• The inclusion of wrist EMG significantly reduces heart rate estimation errors compared to PPG alone.
• Wrist-level EMG electrodes can be unobtrusively integrated into commercial smartwatch form factors without cumbersome forearm straps.
• The approach bridges a crucial reliability gap for consumer fitness tracking, clinical remote patient monitoring, and athletic performance analysis.

@article{friman2022wrist,
  author    = {Friman, Severi and Vehkaoja, Antti and Perez-Macias, Jose Maria},
  title     = {The Use of Wrist EMG Increases the PPG Heart Rate Accuracy in Smartwatches},
  journal   = {IEEE Sensors Journal},
  volume    = {22},
  number    = {24},
  pages     = {24197--24204},
  year      = {2022},
  publisher = {IEEE},
  doi       = {10.1109/JSEN.2022.3219297}
}

Resources & Publication Links

1. Official Publication: IEEE Xplore Digital Library (Document 9942941)
2. DOI Link: https://doi.org/10.1109/JSEN.2022.3219297
3. Author Profile: Jose Maria Perez-Macias (ORCID: 0000-0003-3485-0874)