Google looks to ultrasound to monitor heart rate through ANC headphones
Smartwatches that can monitor a wearer's heart rate use light pulses to detect changes in blood volume. But research from Google proposes using ultrasound from an earphone's speaker driver to look for tiny deviations in the surface of the ear canal.
The method developed by researchers from Google, the University of Princeton and Rutgers University is called audioplethysmography – or APG for short – and involves sending low-intensity ultrasound signals via the speaker drivers of active noise canceling wearables and down the ear canal. The onboard feedback microphones are then used to register the echo that bounces back, and look for minute changes caused by blood flow.
After some mathematical wizardry for clarity, the team noted similarities between the kind of PPG waveforms produced by smartwatch sensors and the APG waveform, with "an improved view of cardiac activities with more pronounced dicrotic notches (i.e., pressure waveforms that provide rich insights about the central artery system, such as blood pressure)."
The ultrasound signal can't be heard by the wearer, and is not affected by music playing through the speaker drivers or poor earbud seals. Early tests did find that body movement could have a negative impact though, but this could be overcome by employing multiple frequencies.
During field tests with 153 participants, the APG solution was found to achieve high accuracy for heart-rate detection (3.21% median error across participants in all activity scenarios) and registering heart rate variability (2.70% median error in inter-beat interval).
The team reckons that any true-wireless active noise canceling headphones can be set to work as smart health monitors after a "simple software update."
Though tempting to think that an APG system might be a new feature in upcoming Pixel Buds, the development will need to jump through some regulatory hoops before that happens so it might be a while before we get to try the system in the wild.
The research was recently presented at MobiCom 2023, and a paper is available via Google Research.