A pillowcase that tracks sleep

A research team at Columbia Engineering has created a unique system called PillowSense that can monitor sleep using a special pillowcase. This innovative system measures heart rate, breathing rate, body movement, and sleep posture through direct contact with the sleeper's neck. The interdisciplinary group of researchers, led by Associate Professor Xia Zhou, worked together to develop this technology, which was presented at the Pervasive and Ubiquitous Computing Conference and received a Distinguished Paper Award.

In over 151 nights of testing in people's homes, PillowSense successfully distinguished seven different sleep postures, achieving an impressive F1 score of 88%. This score measures the system's ability to accurately classify sleep positions. The findings suggest that posture can provide important insights into sleep that vital signs alone cannot capture, such as how sleeping position affects breathing and comfort.

The system works by incorporating conductive fabric into a pillowcase, which acts as an electrode to detect the electrical signals produced by the body during sleep. These signals are then measured using sensitive electronics placed on the sleeper's neck. By separating the heart and muscle signals, the system can estimate heart and breathing rates, as well as identify patterns associated with different sleeping positions.

One interesting discovery made during the research was that neck muscle activity could also indicate whether the sleeper's legs were bent. This finding opened up the possibility of detecting more detailed sleep postures using just a pillowcase. The researchers had to develop advanced signal-processing and machine learning methods to cope with variations caused by changes in contact between the neck and pillow, differences in skin condition, and hair.

Testing the system in people's homes provided valuable insights into how sleep changes over a full night, as well as the impact of movements on the processing required for posture classification. The study included healthy participants, and the researchers found that movement was relatively infrequent, which meant less unnecessary processing. However, poor sensor contact sometimes resulted in gaps in the data, which were excluded from the posture-classification results.

The researchers envision a user-friendly system that people could easily slip over their own pillow and pair with a smartphone app. In the morning, users could review their posture, movement, heart rate, and breathing patterns from the night before. The system was designed with comfort and maintenance in mind, with a detachable sensing layer for hand washing, and the rest of the pillowcase being machine washable.

While PillowSense is currently a research prototype and not yet a diagnostic tool, it has the potential to benefit individuals who find wearable sensors uncomfortable or difficult to tolerate. Additionally, it could help sleep researchers collect observations over many nights in a home setting. The team acknowledges that there are still challenges to address, such as improving sensor contact, addressing interference from hair, simplifying electronics, and initial setup.

A companion app would also need to clearly indicate when measurements are incomplete or uncertain. Ultimately, the goal is to demonstrate that the system remains accurate, comfortable, affordable, and easy to maintain over months of everyday use.