MIT uses radio waves and AI to more accurately study sleep
Sleep tracking has moved to the bedroom, with apps, peripherals and wearables that use movement or your device’s microphone to figure out when you’re sleeping or awake. Those with sleep disorders, however, are still stuck with large, disruptive arrays of physical sensors for sleep studies. Now, however, researchers at MIT have started using radio signals and artificial intelligence algorithms to analyze patients’ sleep stages without physical sensors and they’re reporting a high rate of accuracy. This could help people with Parkinson’s, Alzheimer’s and epilepsy, all of whom can have sleep disruptions that are hard to detect. Eventually, it may help all of us.
While the MIT system is in its infancy now, it’s easy to imagine a near future with home-based sleep monitoring using radio frequencies (RF). “Imagine if your Wi-Fi router knows when you are dreaming, and can monitor whether you are having enough deep sleep, which is necessary for memory consolidation,” said study leader Dina Katabi in a statement. “Our vision is developing health sensors that will disappear into the background and capture physiological signals and important health metrics, without asking the user to change her behavior in any way.”
While other systems use radio signals to monitor sleep, this is the first study that claims a high rate of accuracy (80 percent) as measured against EEG recordings. The RF signals gather some irrelevant information when tracking sleep, so the MIT team had to come up with new algorithms to help separate out the important data. The new sleep monitoring system uses deep neural networks and unique, MIT-written AI algorithms to analyze the data to translate the raw information to valuable sleep data. The team plans to use this new technique to study how Parkinson’s affects sleep next.
The researchers plan to present their research at the International Conference on Machine Learning on August 9. The current sleep monitor builds on previous radio-based systems the team has created that use low-power RF signals to detect and analyze emotions via vital signs like pulse and respiration. They’ve also used RF to measure walking speed, which can help doctors predict cognitive decline, falls and some cardiac or pulmonary diseases.