In order to track how someone's body position changes as they sleep, they typically either have to be rigged up with stick-on sensors, or filmed by a video camera. MIT's new BodyCompass system, however, offers a third alternative that many people may prefer.
First of all, why would a person's sleeping position need to be tracked? Well, for starters, frequent changes in sleep posture are often an indicator of poor sleep quality. And if a sleeping infant doesn't move for a long period of time, it could mean that SIDS (sudden infant death syndrome) is imminent.
Additionally, a Parkinson's disease patient's inability to turn over in bed can indicate that their condition is worsening. And if a person suffers from epilepsy, then they shouldn't sleep on their stomach, as doing so increases their risk of sudden death.
The problem with taping sensors to the skin lies in the fact that the devices are restrictive and often uncomfortable, potentially keeping the patient from sleeping in their usual manner. Cameras are less obtrusive, but they may cause the patient to feel self-conscious, and to worry about where the footage might ultimately be shown.
That's where BodyCompass comes in.
Developed by MIT Computer Science PhD student Shichao Yue and Prof. Dina Katabi, it incorporates a Wi-Fi router-like tool that is mounted on the bedroom wall. That device sends radio signals out into the room, which reflect off of various surfaces, including the person's body. Algorithms on a connected computer can tell which of the reflections come from the sleeper, due to the way that they're affected by the rise and fall of that person's chest as they breath.
By analyzing the paths of these sleeper-specific reflections, the system is able to determine the person's current posture. If they're lying on their back, for instance, then the emitted signals will be reflected off their chest and up to the ceiling, before bouncing off the ceiling and back to the device.
In lab tests, BodyCompass monitored over 200 hours of sleep data from 26 test subjects sleeping in their own bedrooms.
It proved to be 94 percent accurate after being "trained" on a week's worth of sleep data – this involved having the participants wear two strategically-placed motion sensors for that period, so that the system could learn which radio signal reflection paths corresponded to which actual positions. That figure dropped to a still-impressive 87 percent when the training period was shortened to just one night.
It is hoped that once the technology is perfected, it could be paired with a system that would alert sleepers (or infants' parents) when problematic sleep postures or movement patterns are detected.
A paper on the research was recently published in the journal Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies.
Source: MIT