Menu
In the final post in our three-part series about alcohol impairment, we argue that vision-based driver monitoring systems (DMSs) can accurately detect driver impairment, but not BAC.
Blood alcohol concentration (BAC) is the prevailing metric to classify a driver as unfit to drive. However, our research shows that impairment persists long after BAC declines, as outlined in the first article in this series. When we say “impairment”, we refer to the increased likelihood of a person making a mistake while driving by not reacting when they should.
On the back of our research to understand how impairment arises after an elevated BAC, we also developed an eyelid-based model to detect this impairment, which is covered in greater detail in the second article in this series. In the process of developing this model, our data scientists investigated whether visual inputs could also predict a person’s BAC.
Unsurprisingly to our team, visual inputs were much better at measuring impairment than BAC. But what continues to surprise us is how many engineering teams are still attempting to build algorithms within a vision-based DMS using BAC as the ground truth. And they repeatedly struggle to achieve good performance as a result.
This article clarifies why a vision-based DMS is best suited to detecting impairment, but will never achieve high accuracy at measuring BAC.
Working with the team from KEA Technologies in Boston, we conducted a comprehensive study on how impairment develops after alcohol consumption. The BAC and psychomotor impairment of 30 diverse participants were monitored over a 5 to 8-hour period after drinking alcohol. We found that while BAC peaks early and decreases linearly afterwards, impairment continues to increase for hours. In other words, drivers can be severely impaired for hours even when under the legal limit for BAC.
To develop our model, we needed visual data of the participants as they performed the vigilance tests. They were recorded using two cameras in the following locations:
To answer this question, let us revisit the definitions of the two metrics, how they interact, and how they are expressed physiologically.
BAC | Impairment | |
|---|---|---|
What is it? | Amount of alcohol in blood | Increase in risk of driver error |
How is it measured? | Gold standard: Blood Very accurate: Breath | Laboratory: EOO in JTV On-track: Two wheels out of lane twice within 15 minutes |
What DMS sensors can measure it and how? | Breath-based sensors Possibly touch-based sensors (via visible capillaries in the fingers) | Best detected with:
Image-based sensors observing eye and eyelid coordination, as well as head and face movements |
What countermeasures may reduce risk? | Time | Cool air, haptic feedback, audible alerts, ADAS countermeasures |
“Psychomotor impairment can persist for hours after an elevated BAC. This impairment presents as a reduction in coordination of physical movements. Vision-based driver monitoring systems see the subtle changes in a person’s movements, not the level of alcohol in the blood. Accordingly, these systems will always detect impairment far more effectively than BAC.”
Optalert’s research reveals that drivers with BAC levels below the legal limit can still be impaired enough to pose a risk on the road. This has profound implications for road safety, and highlights the need to detect impairment as a separate metric than BAC.
This is why we have developed an algorithm based on eyelid movements to monitor impairment in real-time.
By monitoring eyelid movements via a vision-based DMS, we can accurately detect driver impairment and enact countermeasures.
In such a scenario, there are possible countermeasures to restimulate the driver including:
The advanced driver assistance system (ADAS) can also intervene in a number of ways to reduce the immediate risk from an impaired driver:
The most effective countermeasures remain an active area for further research.
But there is strong evidence that vision-based DMS is best suited for identifying impairment, while BAC measurement should remain the domain of breath and touch-based sensors. Consequently, any algorithm that sits within a vision-based DMS should be both developed and validated against the ground truth of impairment:
Optalert has developed the most accurate impairment detection algorithm due to alcohol. This will revolutionise road safety by accurately identifying impairment in real-time.
With solutions grounded in rigorous research and cutting-edge technology, Optalert’s products provide a proactive approach to driver safety – going beyond traditional approaches to address true fitness to drive based on objective biomarkers.