We are creating some awesome events for you. Kindly bear with us.

U.S.: Deep Learning and Wi-Fi to Detect Breathing Problems

The National Institute of Standards and Technology (NIST) of the United States (U.S.) has created a deep learning algorithm called BreatheSmart. With the help of the Wi-Fi signal, the software can detect when someone in the room is struggling to breathe.

Working with colleagues at the FDA’s Centre for Devices and Radiological Health’s Office of Science and Engineering Labs (OSEL), NIST’s research in shared spectrum metrology is led by Jason Coder and research associate Susanna Mosleh. Together, they developed a new method for measuring a person’s breathing rate using existing Wi-Fi routers. This paper was just published in IEEE Access.

The research project began in 2020 when NIST scientists wanted to assist doctors in combating the COVID-19 pandemic. Patients were isolated, and ventilators were in short supply. Previous research has investigated how Wi-Fi signals can detect people or movements. Still, these setups frequently required custom sensing devices, and the data from these studies were minimal.

“As everyone’s world was turned upside down, several of us at NIST were wondering what we could do to help. We didn’t have time to create a new device, so how can we make do with what we have?” said Coder.

The researchers then generate deep learning software that uses little changes in signal bounce back to Wi-Fi routers to calculate human breathing ability. Wi-Fi routers continuously broadcast radio frequencies to digital devices such as phones, tablets, and computers, it is possible. These devices detect the signal and connect the users to the internet. Invisible frequencies bounce off or pass through everything around them, including walls, furniture, and human movement.

In more detail, the client device (such as a cell phone or laptop) sends the “channel state information,” or CSI, to the access point in Wi-Fi is a set of signals (such as the router). The client device always sends the same CSI signal, and the access point receiving the CSI signals knows what it should be. The CSI signals, however, become distorted as they bounce off objects or lose strength as they travel through the environment. So, the access point analyses the amount of distortion to adjust and optimise the link.

The modifications alter the signal’s path from the router to the devices. As a result, the signal path can detect movements, even as minor as breathing.

Because these CSI streams are small, less than a kilobyte in size, they do not interfere with the data flow over the channel. Therefore, these interactions do not disrupt the internet connection but may indicate that someone is in trouble. The team then modified the router’s firmware to request these CSI streams more frequently, up to 10 times per second, to get a more detailed picture of how the signal was changing.

Furthermore, they set up a manikin outfitted with an anechoic chamber to train medical professionals and a commercial off-the-shelf Wi-Fi router and receiver. This manikin is intended to simulate various breathing conditions, including normal breathing, abnormally slow breathing (called bradypnea), abnormally rapid breathing (tachypnoea), asthma, pneumonia, and chronic obstructive pulmonary disease, or COPD.

When the manikin “breathed,” its chest movement changed the Wi-Fi signal’s path. The team members recorded the data provided by the CSI streams. Despite collecting a wealth of data, they still needed assistance in making sense of what they had gathered.

“This is where we can use deep learning,” Coder explained.

Deep learning is a subset of artificial intelligence, a type of machine learning that mimics humans’ ability to learn from their previous actions and improves the machine’s ability to recognise patterns and analyse new data.

Mosleh worked on a deep learning algorithm to sift through the CSI data, comprehend it, and identify patterns that indicated various breathing problems. As a result, the BreatheSmart algorithm successfully classified different respiratory patterns simulated with the manikin 99.54 per cent of the time. Coder and Mosleh hope that app and software developers can use the work process as a framework to create programmes to monitor breathing remotely.

Send this to a friend