A new research paper from Apple’s Machine Learning Research section of Cornell University’s arxiv.org shows that respiratory rates can be monitored using AirPods or other audio wearable technology using breath audio.
According to the paper, which was first reported on by MyHealthyApple, researchers discovered that this could create a cost-effective way of tracking disease progression and cardio-respiratory fitness over time while using “accessible, aesthetically acceptable”, devices.
Data was collected from 21 individuals using microphone-enabled, near-field headphones before, during, and after strenuous exercise.
RR was manually annotated by counting audibly perceived inhalations and exhalations.
A multi-level convolutional neural network was used to achieve signal clarity among other things and the results observed show that RR can be estimated with a concordance correlation coefficient (CCC) of 0.76 and a mean squared error (MSE) of 0.2, demonstrating that audio can be a viable signal for passively estimating RR.
As the paper points outs, the respiratory rates aren’t 100 percent accurate. However, having access to a remote estimate of the data. Typically, those with conditions in which respiratory rates must be measured have to interact with their health care provider on a somewhat regular basis. This could prove to provide alternative measures.
It’s also worth pointing out that the paper doesn’t specifically call out which device was used on the 21 participating individuals. Theoretically, it very well could have been based around the use of a pair of AirPods. In which case, we may see a future generation of AirPods utilize better near-field microphone technology.