Paper: Smart Homes that Monitor Breathing and Heart Rate [PDF]
Authors: Fadel Adib, Hongzi Mao, Zachary Kabelac, Dina Katabi, Robert C. Miller
Tags: New System, Smart Home, Sensors, Health & Wellness, Bio-sensing, Internet of Things,
Fadel Adib and his colleagues work in the Wireless Lab at MIT (under Dr. Robert Miller). Most of their research utilizes different radio waves and using their reflection to determine the presence of people in an indoor space. But this year, they produced a much higher resolution — Vital Radio — that records respiration rate and heart rate. This is a huge advance for the internet of things and smart homes. Look ma’ no wires!
"In this paper, we ask whether it’s possible for smart homes to monitor our vital signs remotely – i.e., without requiring any physical contact with our bodies.” People can just relax in their homes — note that the technology used here does require they sit or lie down — as they normally would and the system generates highly accurate respiration rate and heart rate data.
The system is highly accurate, with average accuracy rates over 98%. The researchers used consumer grade chest-strap heart rate monitors as their ‘ground truth’ measurements. And they can do this at a distance of almost 25 feet (8m)!
The system works through a property that the same researchers discovered in 2013, Frequency Modulated Carrier Waves (FMCW). At a simplified level, Adib created an algorithm to segment the reflections into a variety of bins, and then focus in on bins that have movement in them (i.e., a chair or desk will reflect consistently, but a human body is always in slight motion).
The FMCW is robust to two kinds of interference:
- First, body motions or limb movements can cause the sensitive reflections to be lost in the noise.
- Second, two different individuals in the space could also cause confusion since both will be measured simultaneously.
For the first scenario, "Vital-Radio exploits that motion due to vital signs is periodic, while body or limb motion is aperiodic. It uses this property to identify intervals of time where a user’s whole body moves or where she performs large limb movements and discards them so that they do not create errors in estimating vital signs.” The current implementation uses a 30-second window and if the data isn’t periodic enough, the system just discards that data.
To answer the second kind of interference, the experimenters measured up to three individuals in the same room. Because system only calculates data from a single bin at a time, collisions are much less frequent. But when the individual bodies are very close together, interference does appear and accuracy does drop to approximately 75%.
What can you learn from this project:
- The possibility of sensing detailed medical data without any contact or traditional "sensors"
- The powerful idea to build systems using leftovers or "noise" from other systems. Adib did similar work using WIFI signals to detect presence and motion indoors too (see the youtube video here, see an image below). The idea of using other people's noise as your signal is a hallmark of advanced ubiquitous computing systems.
- Good ways to think about designing experiments to measure the effectiveness of proof-of-concept technologies. In cases where you can get ground truth data, comparative evaluations are helpful.