We all get sleepy at work from time to time, especially after a big lunch. But for people whose jobs involve driving or working with heavy machinery, drowsiness can be extremely dangerous, if not downright deadly. Drowsy driving contributes to hundreds of fatal traffic accidents in the U.S. each year, and the National Safety Council has cited drowsiness as a critical risk in construction and mining.
To help protect drivers and machine operators from the dangers of falling asleep, engineers at the University of California, Berkeley, have created prototype headphones that can detect signs of drowsiness in the brain.
The headphones detect brain waves in the same way as an electroencephalogram (EEG), a test doctors use to measure the brain’s electrical activity. While most EEGs detect brain waves using a series of electrodes attached to the head, the headphones do so using built-in electrodes that are designed to come into contact with the ear canal.
The electrical signals detected by the headset are smaller than those picked up by a traditional electroencephalogram. But in a new study, researchers show that their in-ear EEG platform is sensitive enough to detect alpha waves, a pattern of brain activity that increases when you close your eyes or start to fall asleep.
“I was inspired when I bought my first pair of Apple AirPods in 2017. I immediately thought, ‘What an incredible platform for neural recording,’” said senior study author Rikky Muller, an associate professor of electrical engineering and computer science at UC Berkeley. “We think this technology has many potential uses and that classifying sleepiness is a good indicator that the technology can be used to classify sleep and even diagnose sleep disorders.”
Using an earpiece as an EEG electrode poses several practical challenges. To obtain an accurate EEG, the electrodes must make good contact with the skin. This is relatively easy to achieve in traditional EEGs, which use flat metal electrodes attached to the scalp. However, it is much more complicated to design an earpiece that fits comfortably and snugly on a wide variety of ear sizes and shapes.
When Muller’s team started working on the project, other groups developing in-ear EEG platforms were using wet electrode gels to ensure a good seal between the earpiece and the ear canal or creating custom-molded earpieces for each user. She and her team wanted to design a model that was dry and user-generic, so anyone could put it in their ears and get reliable readings.
“My personal goal was to try to create a device that could be used every day by someone who would really benefit from it,” said Ryan Kaveh, a postdoctoral researcher at UC Berkeley and co-lead author of the study. “To accomplish that, I knew it would have to be reusable, adaptable to a variety of people, and [be] “Easy to manufacture.”
Kaveh led the study with graduate student Carolyn Schwendeman and collaborated with Ana Arias’ lab at UC Berkeley to design the final earpiece in three sizes: small, medium and large. The earpiece incorporates multiple electrodes in a cantilever design that applies gentle outward pressure into the ear canal and uses flexible electronics to ensure a comfortable fit. Signals are read through a custom, low-power wireless electronic interface.
In a 2020 paper, researchers showed that these headphones can detect a range of physiological signals, including eye blinks, alpha brain waves, and the auditory steady-state response, which is the brain’s response to hearing a constant tone. In the new study, they improved the design of the headphones and incorporated machine learning to demonstrate how they could be used in a real-world application.
As part of the experiment, nine volunteers were asked to wear the headphones while performing a series of boring tasks in a darkened room. From time to time, the volunteers were asked to rate their level of sleepiness and their response times were measured.
“We found that even when the signal quality from the headphones seemed worse, we were still able to classify the onset of drowsiness with the same level of accuracy as much more complicated and bulky systems,” Kaveh said. The headphones also retain their accuracy when categorizing drowsiness in first-time users, a characteristic of devices that can work “out of the box.”
Muller, who developed the Ear EEG with support from the Bakar Fellowship and the Bakar Prize, is continuing to refine the design and explore other potential applications for the device, which can also record signals beyond EEG, such as heartbeats, eye movements and jaw clenching.
“Wireless earbuds are something we already wear all the time,” Muller said. “That’s what makes Ear EEG such an attractive option for wearables. It doesn’t require anything extra.”
This study was funded in part by the Ford University Research Program and a Bakar Spark Award.