These Earbuds Know When You're Drowsy

When applied to commercial applications, wearable electronic devices have the potential to unlock new levels of productivity and safety. We have only just begun to scratch the surface of what is possible, but one of the early applications to emerge is drowsiness detection. When going through a regular routine, we can all go on autopilot and start to feel a bit sleepy at work from time to time. But when that job happens to involve long-haul driving or operating heavy machinery, that drowsiness is a hazard that cannot be ignored for a moment.

If that drowsiness results in the operator closing their eyes for even a few seconds, it can lead to a disastrous situation. Sounding an alarm of some sort can bring awareness to them such that they can get off the road or take other measures to raise their level of alertness. In order to do that, some system first needs to be in place that can immediately recognize the signs of drowsiness.

One of the most reliable ways of telling when someone is feeling sleepy involves monitoring the electrical activity in their brain. This can be accomplished noninvasively by performing an electroencephalogram (EEG) test to look for certain characteristic patterns. However, EEGs require that an array of electrodes be attached to the head, which is not exactly desirable or practical for use in the real world.

A team at the University of California Berkeley found that similar data could be collected in a much more compact device, which could make the technology much more available. Using a pair of earbuds, much like the ones many of us already frequently wear, they demonstrated that it is possible to capture the electrical signals necessary to reliably detect drowsiness.

The researchers are by no means the first to capture the brain’s electrical activity from earbud-like devices. But in the past, these devices relied on either custom-molded earpieces or wet electrode gels to get a good enough contact with the skin to capture high-quality measurements. These solutions are either expensive or cumbersome, and so are not conducive to being used widely.

To overcome these present limitations, the team designed a new type of earpiece. They are not quite one-size-fits-all devices, but instead come in small, medium and large sizes, which is not unreasonable from a manufacturing standpoint. That got the team part of the way there, but not quite to the tight contacts needed for proper sensing operation. To make good contact, the earpieces also incorporate a cantilevered design that applies gentle pressure to the electrodes, ensuring good contact with the skin without compromising comfort or practicality.

Experiments revealed that the signal quality was nowhere near as good as what could be obtained with a traditional EEG test, which is not really surprising. However, when combined with a machine learning classification algorithm, the earpieces were quite capable of detecting alpha waves, which are strongly associated with drowsiness. In fact, in a series of trials it was shown that drowsiness could be detected with the same level of accuracy as a bulky, traditional EEG instrument.

Looking ahead, the team intends to work toward running the machine learning algorithm on-device. This step would go a long way toward making the device practical for real-world use.