Typical heart rate monitoring in wearable tech, like smart watches or wireless earbuds, relies at least partially on photoplethysmography (PPG), which uses light pulses to measure blood activity. It works generally well, but it has its limitations. Google scientists wrote in a new research blog spotted by 9to5Google yesterday that they had tried a different approach, called audioplethysmography (APG), that uses ultrasound to measure heart rate. And they did it with off-the-shelf active noise-canceling (ANC) earbuds and a software update.

The trick works by bouncing a low-intensity ultrasound signal off the inside of the ear canal and using the tiny microphone that helps make ANC work to detect skin surface perturbations as blood pumps through it. According to the blog, the technique was “resilient” even given a bad ear seal, differing ear canal size, or darker skin tones. That last one is notable since heart rate accuracy with darker natural skin tones or tattoos has been an ongoing problem with smartwatches and other wearables until now.

Google’s researchers also found the ultrasound approach worked fine when music was playing, but said that it had issues in noisy environments and that “the APG signal can sometimes be very noisy and could be heavily disturbed by body motion.” However, they found they could overcome the motion problem by using multiple frequencies and teasing out the most accurate signal among them.

In addition to commercially available earbuds, the researchers also used purpose-built prototypes to test the effect of microphone placement. The field study was performed with 153 participants. The researchers said the median error rate for heart rate and heart rate variability was 3.21 percent and 2.70 percent, respectively.