A mattress cover that tracks sleep, with its recipe in the open
Researchers in Pisa built a sensor-lined mattress cover and published the full method for turning its signals into sleep estimates. In 13 people at home, it matched a wrist actigraph well on total sleep time and less well on how broken the night was.
Soma Labs3 min read
Most sleep trackers ask you to wear something. Rings, watches and chest straps all work, but some people take them off, forget to charge them or simply sleep better without them. A growing group of devices goes the other way and hides the sensors in the bed.
A team in Pisa has now tested one of these, a smart mattress cover, in people's own homes. Just as notable, they published exactly how the cover's raw signals are turned into sleep numbers, something most commercial in-bed trackers keep private.
The setup
Eleonora Melissa and Marco Laurino of the Institute of Clinical Physiology at Italy's National Research Council (CNR), with Carlotta Marinai, Nicola Carbonaro and Alessandro Tognetti of the University of Pisa, built the cover as part of TOLIFE, a European project on long-term monitoring at home.
The cover holds a 40-point textile pressure grid under the chest area and two motion sensors (accelerometers), zipped inside a removable sleeve. A small hub sends the data to the cloud, and a bedroom box logs light, sound, temperature, humidity and air quality.
Thirteen healthy adults aged 24 to 38, six women and seven men, slept on it at home for 3 to 13 nights each, with no rules about bedtimes or routines. Each also wore a research-grade wrist actigraph, the reference for the comparison. Of 60 recorded nights, 55 were usable.
The team adapted a classic actigraphy scoring method (Cole-Kripke) to use both pressure and motion, and let the wake threshold adjust to each person and night. They built it on 18 nights from 8 people and checked it on the remaining 37 nights, including 5 people the model had never seen.
What the study found
Total sleep time: good agreement. Against the actigraph, the cover's estimate of total sleep time correlated at r = 0.71, with an agreement score (ICC) of 0.74. On average it was off by about 37 minutes a night in either direction.
Number of awakenings: moderate. Agreement was 0.55. Here the pressure grid added the most, picking up shifts that motion alone missed.
Time awake and sleep efficiency: weak. Agreement on minutes awake after falling asleep was 0.22, and on sleep efficiency 0.29. The differences also grew as nights got more broken.
What to keep in mind
Small and young. Thirteen healthy adults under 40 is a first look, not a verdict. The authors say larger and more varied groups are needed.
The reference is not perfect. A wrist actigraph estimates sleep from movement too. The standard for sleep studies is a full overnight lab recording (polysomnography), which this study did not use.
Beds differ. Mattress type, bedding and sleeping position can change how pressure and movement reach the sensors.
It's a prototype. The cover was built for a research project and isn't a consumer product. The work was funded by the EU's Horizon Europe programme, and the authors declared no commercial conflicts.
Where Soma fits
Soma doesn't make a mattress sensor, and the Soma app, which is not yet available, doesn't estimate sleep from your bed. What this study shows is broader: duration is the easiest sleep number to get right, and the finer details of a broken night are much harder for any single sensor. Read those finer numbers as a trend to watch over weeks, not a nightly grade.
Sources, as of November 9, 2026: Eleonora Melissa, Carlotta Marinai, Nicola Carbonaro, Alessandro Tognetti and Marco Laurino, "The smart mattress cover system for unobtrusive home sleep monitoring and automated sleep quality assessment," Frontiers in Digital Health, doi 10.3389/fdgth.2026.1958146, published under CC BY 4.0. This post describes research and isn't medical advice.