According to the report, Apple is further studying health monitoring based on wearable devices and detailed its commitment to estimate respiratory rate with the help of AirPods in a research paper published on Wednesday. My Healthy Apple published an article titled Estimating Respiratory Rate From Breath Audio Obtained Through Wearable Microphones on Apple’s machine learning research website.
It introduces a method of using audio data collected from AirPods to monitor the breathing rate of healthy people during exercise. Apple hopes to prove that ready-made, beautiful and relatively inexpensive devices like AirPods can be used to estimate breathing rate and track cardiorespiratory health.
Although this article did not specify the AirPods model, it clearly pointed out that the breathing audio collected from the microphone of the wearable device is used to inform a learning network model, with the purpose of distinguishing normal and heavy breathing. According to the study, breathing frequency can also be estimated by detecting breathing sound patterns.
Although sensors such as thermistors, respirometer sensors, and sound wave sensors can most accurately estimate a person’s breathing pattern, they are invasive and may not be suitable for daily use. In contrast, wearable headsets are relatively economical and convenient, Comfortable and beautiful.
Apple’s research focuses on estimating the breathing rate during physical activity, but the researchers point out that similar techniques may also be applied to clinical situations related to dyspnea. Dyspnea during exercise is often used in medical research, and it may be a powerful independent predictor of mortality.
In the process of collecting data, Apple asked test participants to record a series of audio clips before, during and after exercise. The accompanying data includes heart rate readings from Apple Watch. The data is parsed and analyzed with the help of convolutional neural networks to represent the individual’s breathing rate.
The process includes detecting and mitigating incidents of background noise. Apple concluded that the system can achieve a consistent correlation coefficient (CCC) of 0.76 and a mean square error (MSE) of 0.2. These indicators are considered feasible indicators for estimating respiratory rate.
As far as we know, no previous studies have investigated the data collected from the natural conditions of indoor and outdoor background conditions, using perceptual classification data and trying to build an end-to-end system. These will consume the energy of the filter library, which can be directly predicted. Respiration rate and heavy breathing classification, Apple said.
Whether Apple intends to apply AirPods-based respiratory rate detection to its health kit based on this discovery is still unknown. There are rumors that future wearable devices will include health monitoring sensors similar to Apple Watch hardware, but it is unclear whether and when these models will be released.
