NIST’s Wi-Fi System Pinpoints People Struggling to Breathe
NIST researchers recently developed the BreatheSmart algorithm, which detects abnormal respiration without using any new hardware.
Aiming to create a ubiquitous health monitoring mechanism, the National Institute of Standards and Technology (NIST) has developed a "BreatheSmart" algorithm that uses Wi-Fi signals to monitor respiration wirelessly. NIST scientists claim the COVID-19 pandemic motivated their research to develop a device that monitors respiratory health without requiring complex new hardware.
Although NIST’s BreatheSmart is not the only mechanism through which respiration can be monitored, it does present major benefits compared to other methods. Since it makes use of the Wi-Fi standard’s channel state information, it can be deployed exclusively in software without requiring any additional hardware.
The system architecture behind the BreatheSmart algorithm highlights the ability to identify breathing patterns without requiring new hardware. Image used courtesy of NIST
In this article, we'll examine the BreatheSmart algorithm alongside some other methods of wireless breath monitoring to determine its benefits and tradeoffs as a noninvasive respiratory health metric. We'll also discuss the future of NIST's technology to assess how it can be integrated into day-to-day life.
Two Features in One
Channel state information (CSI) helps to compensate for the environment in which the Wi-Fi router is deployed. The CSI includes information about reflections, attenuation, and path length changes created by environmental changes to allow the transmitted and received signals to be normalized and read appropriately.
The test setup for the BreatheSmart algorithm uses only two Wi-Fi devices, the access point and the client, to characterize the target’s motion using the CSI. Image used courtesy of NIST
In addition to protecting the integrity of Wi-Fi signals, this feature encodes small environmental changes caused by biological motion. Human breathing causes small chest movements that will alter the signal path from transmitter to receiver and as such, will be encoded in the CSI stored in the Wi-Fi access point. With an appropriate algorithm such as BreatheSmart, this information can be used to determine respiration rates and identify troublesome breathing patterns.
Using Wi-Fi to Monitor Breathing
The Wi-Fi hardware on its own is not enough to detect problems with breathing. To identify and characterize abnormal breathing patterns, a deep learning model is used on the CSI. After preprocessing, the data can be fed into the deep learning model after training and testing to effectively characterize the observed breathing patterns.
To train and test the model, NIST used a “RespiPro” manikin. The manikin included a realistic airway and programmable breath typically used to train medical professionals. In this application, however, it was used to train the deep learning model.
The test setup using the RespiPro manikin helped researchers characterize the BreatheSmart algorithm’s effectiveness to identify breathing patterns. Image used courtesy of NIST
After training, initial tests with the BreatheSmart algorithm and RespiPro manikin showed a 99.54% success rate in identifying the breathing patterns of the manikin. This measure is, of course, affected by various parameters such as frames per second and attenuation, but still presents a successful initial test for measuring biological motion using existing hardware.
Getting a Nonstop View of Your Health
The algorithm from NIST isn’t the only method of detecting respiration. Techniques such as UWB radar, optics, or capacitive sensing all provide similar abilities to detect small physiological motions, but each comes with tradeoffs. Compared to BreatheSmart, each requires additional sets of hardware to function properly.
While there is no one-size-fits-all solution to health monitoring, BreatheSmart shows promise as an inexpensive, non-invasive way to monitor respiration. Researchers are still developing and testing algorithms to help identify weak points or potential improvements. For example, researchers have considered CSIR as a more robust alternative for measuring respiration than CSI.