This new deep learning algorithm uses Wi-Fi signals to identify breathing troubles

How does a router analyze an individual's troubled breathing pattern?

That's when Coder and research associate Susanna Mosleh devised a way to use existing Wi-Fi routers to measure the breathing rate of a person in the room. 

According to the release, the CSI (channel state information) signal sent from the client device (such as a cellphone or laptop) to the access point (such as the router) is always the same, and the access point knows what the signal should look like.

But, as the CSI signals travel through the environment, they bounce off things or lose strength and get distorted. When this happens, the access point analyzes the amount of distortion to adjust and optimize the link.

As the CSI streams are small, less than a kilobyte, it doesn't hinder the flow of data over the channel. And the team modified the firmware on the router and received CSI streams more frequently, up to 10 times per second, thereby receiving a clear picture of the changing signal.

Leveraging deep learning to make sense

A manikin used to train medical professionals was set up in an anechoic chamber with an off-the-shelf Wi-Fi router and receiver. The manikin replicated several breathing conditions, from normal respiration to abnormally slow breathing, abnormally rapid breathing, asthma, pneumonia, and chronic obstructive pulmonary disease, or COPD. 

As the manikin breathed, the movement of its chest altered the path traveled by the Wi-Fi signal. The team members collected the data sent by the CSI streams. But, even though they had a wealth of data, they needed to make sense of what they gathered. 

"This is where we can leverage deep learning," Coder said.

The algorithm successfully classified respiratory patterns

Mosleh worked on a deep learning algorithm to "comb through the CSI data, understand it, and recognize patterns that indicated different breathing problems." BreatheSmart successfully classified a variety of respiratory patterns simulated with the manikin 99.54 percent of the time, according to the release.

"Most of the work that's been done before was working with very limited data," Mosleh says. "We were able to collect data with a lot of simulated respiratory scenarios, which contributes to the diversity of the training set that was available to the algorithm."

Coder and Mosleh hope that developers can use the process presented in the work as a "framework" to create programs to remotely monitor breathing. 

"All the ways we're gathering the data is done on software on the access point (in this case, the router), which could be done by an app on a phone," Coder said. "This work tries to lay out how somebody can develop and test their algorithm. This is a framework to help them get relevant information."

Their work was published in IEEE Access.

Study Abstract:

Respiratory motion (i.e., motion pattern and rate) can provide valuable information for many medical situations. This information may help in the diagnosis of different health disorders and diseases. Wi-Fi-based respiratory monitoring schemes utilizing commercial off-the-shelf (COTS) devices can provide contactless, low-cost, simple, and scalable respiratory monitoring without requiring specialized hardware. Despite intense research efforts, an in-depth investigation on how to evaluate this type of technology is missing. We demonstrated and assessed the feasibility of monitoring and extracting human respiratory motion from Wi-Fi channel state information (CSI) data. This demonstration involves implementing an end-to-end system for a COTS-based hardware platform, control software, data acquisition, and a proposed processing algorithm. The processing algorithm is a novel deep-learning-based approach that exploits small changes in both CSI amplitude and phase information to learn high-level abstractions of breathing-induced chest movements and to reveal the unique characteristics of their difference. We also conducted extensive laboratory experiments demonstrating an assessment technique that can be replicated when quantifying the performance of similar systems. The results indicate that the proposed scheme can classify respiratory patterns and rates with an accuracy of 99.54% and 98.69%, respectively, in moderately degraded RF channels. Comprehensive data acquisition revealed the capability of the proposed system in detecting and classifying respiratory motions. Understanding the feasible limits and potential failure factors of Wi-Fi CSI-based respiratory monitoring scheme—and how to evaluate them—is an essential step toward the practical deployment of this technology. This study discusses ideas for further expansion of this technology.