Their approach uses an advanced artificial intelligence (AI) algorithm to analyze the radio signals around a person and then translate that data into light, deep, or rapid eye movement (REM) sleep stages. This new method could make it easier to diagnose and study sleep problems, say the researchers, as current methods require attaching electrodes and a variety of other sensors to patients and tethering them to monitors such as electroencephalography (EEG) machines in a sleep lab.
Their system consist of a laptop-sized wireless device that produces low-power radio-frequency (RF) signals that reflect off of a sleeping person. Any slight movement by the person changes the frequency of the reflected waves, producing a "signature" that the body leaves on the RF signal, which can then be analyzed for vital signs such as pulse and breathing rate.
"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," says Dina Katabi, the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science at MIT, who led the study. "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.”
In order to translate the measurements of pulse, breathing rate, and movement into sleep stages, the researchers had to develop a new AI algorithm based on deep neural networks to eliminate irrelevant information in the data, such as signals that bounce off of other objects in the room. The resulting algorithm, they say, can be used in different locations and with different people, without any calibration, and was found in testing to be about 80% accurate - comparable to ratings determined by sleep specialists based on EEG measurements.
The researchers next plan to use the technology to study how Parkinson's disease affects sleep. Other potential uses could