Written by Dr Nikolas Williams, EMOTIV Research Scientist.
Several months ago I moved back to the US after eight years of living overseas. Part of starting over again meant purchasing all the things one needs for life. In addition to a couch, bed, and dining table, I of course needed a car. Considering myself a financially savvy person, I looked exclusively for older, cost-effective models but was quickly demoralized by the inflated prices and scarce inventory. The used car market of 2021 was effectively bullying me into buying new, which I ultimately did. My dismay over violating basic personal finance tenets was quickly replaced with unbridled enthusiasm for the comfort and features that came with my brand new Toyota SUV.
I was particularly taken by the autonomous driving features that, to this point, I had only read about. Assisted steering and forward-looking radar made long drives a breeze. I merely had to keep my eyes on the road and a hand resting on the steering wheel and my car basically drove itself. Factor in the collision-avoidance, blind-spot monitoring, rear-facing cameras with an alert system to make sure I didn’t back into anybody crossing behind me, and this new car was objectively orders of magnitude safer than the older model cars I had driven for the better part of the last decade.
Cars, of course, don’t yet drive themselves. While they have nifty autonomous and safety features, cars still require driver supervision and, when necessary, intervention. We are a long way from removing the human component from driving and it is this component that is predominantly responsible for automobile accidents and fatalities. Humans mess up behind the wheel. Whether they decide operating a vehicle after drinking is a good idea, or that speeding is fun, or that they need to eke out just a couple more miles before they pull over to rest their tired selves, humans cause a lot of preventable automobile incidents.
According to the National Highway Traffic Safety Administration (NHTSA), there were 36,096 motor vehicle traffic fatalities in 2019. For 2020, fatalities are estimated to be over 38,000 . A large percentage of these are due to risky driving and thus preventable. The NHTSA has identified six types of risky driving: Speeding, drunk and drug-impaired driving, not using (or improperly using) seat belts, distracted driving, and drowsy driving. As two-thirds of all traffic fatalities can be attributed to speeding and impaired driving, many intervention campaigns are rightly aimed at addressing these risks. However, distracted and drowsy driving result in a non-trivial number of fatalities with 3,142 distraction-related deaths and 697 drowsiness-related deaths in 2019 .
Using neuroscience to measure attentiveness in the lab
Neuroscientists use various methods to measure attention in the laboratory. One of these methods capitalizes on the fact that our brain releases tiny amounts of electricity as its neurons fire. Using electroencephologram (EEG), we can measure the fluctuations in this electricity to understand when and where the brain is active. The speed, or frequency, at which these fluctuations occur are known as oscillations, or more commonly, brain waves. The frequency of brain waves can provide insight into mental states or processes.
For example, brainwaves that oscillate 14 to 30 times per second (or 14 – 30 Hz) are known as beta waves and are associated with high levels of mental engagement. Oscillations in the 8 – 13 Hz range are known as alpha waves and are generally present during periods of relaxation or passive attention. For example, you would often see alpha waves when a person was meditating. Theta waves are oscillations between 4 and 7 Hz and seen when a person is deeply relaxed or drowsy. The slowest waves are delta waves (1 – 4 Hz) and are observed when a person is deeply sleeping.
See related post The Introductory Guide to EEG
In the lab, scientists can measure the timing, magnitude, and frequency of brain waves to determine how engaged or disengaged a person’s mind is during tasks. For example, when a person sees or hears something for which they have been watching, their EEG shows a very specific response called a P300, which is a large amplitude wave that occurs about 300 ms after the appearance of the object . Likewise, a decrease in alpha oscillations can indicate someone is paying close attention to something . Being drowsy also produces detectable EEG signatures by way of changes in delta, theta, and alpha oscillations .
How can we measure attentiveness in a car?
In a vehicle we can measure attentiveness and drowsiness using behavioral methods. For example, cameras could track drivers’ eyes to make sure that they are looking at the road. Likewise, cameras could detect when drivers’ heads begin to nod indicating that they are sleepy. However, just because a person is looking at the road or their head is not drooping does not mean they are paying attention or that they are not fatigued. EEG can augment the detection of these hazardous states. They may even be able to predict them before they are behaviourally detectable.
In 2020, researchers conducted a systematic review of studies that used commercially available EEG headsets to detect real-time drowsiness . They reported that the most used headset in these types of studies were those manufactured by EMOTIV, followed by Neurosky, Interaxon, and OpenBCI. For drowsiness detection, they found that even basic EEG features, such as frequency oscillations, could be used to detect drowsiness. However, they note that in many cases, “algorithmic optimization remains necessary”, meaning that machine-learning algorithms resulted in more accurate detections.
Leveraging commercial EEG and machine learning algorithms to help keep us safer
EMOTIV has been the leader in commercial EEG for over a decade. During this time they have developed EEG systems in various forms, from 32-channel traditional research caps to 2-channel in-ear headphones. Systems with compact form factors, like the MN8 headphones or Insight, represent the first steps toward everyday, weable neurotech. By integrating these types of hardware into automobile controls, we may be able to prevent accidents before the contributing mental states ever occur.
Integrating EEG hardware into vehicles is only part of the solution. In order to capitalize on acquired brain data, we need to process it into useful metrics. Sophisticated machine-learning algorithms achieve this by decoding EEG data into features that can index specific mental states. To date, EMOTIV has developed seven such detections: frustration, interest, relaxation, engagement, excitement, attention, and stress. EMOTIV engineers have worked closely with neuroscientists to develop these detections through rigorous experimental studies that use protocols known to elicit these states. In the automobile domain, Emotiv is currently fine-tuning a driver-distraction detection developed within a driving simulator. This follows promising results from a collaboration with the Royal Automobile Club of Western Australia, which resulted in an attention-powered car that slowed down when attention waned . You can find some videos of the collaboration and the results on YouTube.
Neuroscience and the future of driving
From early interventions such as seat belts and rumble strips to modern ones such as automatic emergency braking and assisted steering, our cars have become much safer. Yet the number of people who die from accidents each year demonstrates that we have a long way to go before we reach the point where vehicles can be considered “safe”. As technology advances, our cars will no doubt continue to get safer, but as long as humans are the predominant vehicle operators, there will continue to be human-caused crashes. EEG technology represents a particularly promising avenue for mitigating the human factor by detecting subtle indicators and intervening before accident-causing conditions ever occur.
 National Center for Statistics and Analysis, “Early estimate of motor vehicle traffic fatalities in 2020.” National Highway Traffic Safety Administration, May 2021. Accessed: Jan. 04, 2022. [Online]. Available: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/813115
 G. Thut, “Alpha-Band Electroencephalographic Activity over Occipital Cortex Indexes Visuospatial Attention Bias and Predicts Visual Target Detection,” J. Neurosci., vol. 26, no. 37, pp. 9494–9502, Sep. 2006, doi: 10.1523/JNEUROSCI.0875-06.2006.
 C.-H. Chuang, C.-S. Huang, L.-W. Ko, and C.-T. Lin, “An EEG-based perceptual function integration network for application to drowsy driving,” Knowl.-Based Syst., vol. 80, pp. 143–152, May 2015, doi: 10.1016/j.knosys.2015.01.007.
 J. LaRocco, M. D. Le, and D.-G. Paeng, “A Systemic Review of Available Low-Cost EEG Headsets Used for Drowsiness Detection,” Front. Neuroinformatics, vol. 14, p. 42, 2020, doi: 10.3389/fninf.2020.553352. “Australia researchers unveil ‘attention-powered’ car,” 2013. https://phys.org/news/2013-09-australia-unveil-attention-powered-car.html (accessed Jan. 12, 2022).