Alex Moss, CEO of Canaria Technologies, discusses how predictive biometrics could revolutionize mine workforce safety and productivity
By Carly Leonida, European Editor
In September, Macarthur Minerals announced it had signed an agreement with Canaria Technologies to introduce and test the new Canaria-V predictive biometrics platform with a view to wider implementation at its Lake Giles iron ore project in the Yilgarn region of Western Australia during 2021.
Canaria’s system is the first in heavy industry to utilize predictive biometric technology to identify and prevent potential safety-based incidents on-site, but it could, eventually, have wider applications.
“Predictive biometrics include any type of technology that uses historical biometric data from users to identity patterns in that data, then take a next step to make predictions about how that data will behave in the future,” Alex Moss, CEO of Canaria Technologies, explained.
“It’s hard to say exactly where the technology originated, it’s an emerging field that builds upon biometric sciences and academia. We’re one of the first companies to use it outside of laboratories and research institutes.”
Leaders in the field of predictive biometrics include Professor David Clifton’s work at the University of Oxford and various studies at MIT. Most leading universities today have a professor whose work touches upon the subject, even if they’re not fully researching it.
“Interestingly, the techniques used to create time-series-based predictive biometric systems have been used in the financial sector for the past decade and are now being applied to the medical sciences,” Moss added. “But, at Canaria, we find ourselves recruiting more full-stack developers and data scientists, because very few people have worked in predictive biometrics before.
“My company’s technology originated from a project for NASA in 2016 for monitoring astronauts remotely, so I suppose it wouldn’t be wrong to say that it came from space.”
What is ‘Biometric’ Data?
Any data that is gathered about the way the human body functions can be described as biometric.
The best known types are: fingerprint scans; iris scans; electroencephalography (EEG) readings, taken using skullcaps in hospitals; electrocardiogram (ECG) readings where pads are placed on the chest to measure electrical heart activity; and pulse plethysmography (PPG), which measures oxygen levels in the blood.
In early December, Canaria’s work culminated in the launch of the Canaria-V platform, the successor to the Puck, which is currently in the process of being decommissioned. The technology is delivered via a multitiered subscription-based service that includes hardware (a wireless earpiece for each user and smart charging docks that can hold multiple devices) as well as software and cloud-based analytics to support operator safety in mines.
“At Canaria, we use a few different types of biometric data simultaneously to increase the accuracy of our systems and lay down foundations for future use cases,” Moss explained. “We use a Linux-style of development for building our hardware. The sensor systems are, essentially, overengineered and some of the capabilities left dormant. When ready, these can be activated via a software update. For instance, today, the system can provide heat exhaustion predictions, cognitive fatigue predictions and man down alerts, but asphyxiation prediction may be added to that in the near future.”
The Canaria-V Earpieces are themselves miniature computers with their own processing power.
“We lighten this processing load by offsetting segments of the data that our devices collect on to smart docks and then finally through our cloud-computing system,” explained Moss. “The really heavy data analytics work is done on powerful computers by our data scientists and we present the data in a few ways, such as smart phones linked to our devices and on our desktop app, too. We’re working on integrating this into our client’s most used management software, so that users aren’t overwhelmed with lots of different platforms.”
Because the devices themselves have edge processing power, they can handle data and issue critical alerts both with and without an internet connection — a crucial feature at remote or underground sites — but having cloud power as well means that more complex sets of raw data can be computed and stored for future potential applications.
“For instance, we use PPG and heart rate variability measurements to predict cognitive fatigue today, but there could be parts of that data that we don’t understand yet that in five-years’ time could be a crucial building block for another application,” Moss said. “It’s important that we have that data, so we’ve built a few different layers of storage contingency on our devices. The earpiece itself is a microcomputer with on-device storage capabilities, the charging docks offer additional storage as contingency, and then we also have a backup cloud storage.”
The team has found that, in practice, some mine sites can go for five days without a reliable internet connection, so the earpieces have been engineered to store data on-device to mitigate this data-transfer risk. Once on the charging dock, the data is automatically downloaded to clear the device and allow it to be pushed up to cloud via an ethernet connection. The previous generation’s dock can store up to 10 days’ worth of data from each device in case of poor connectivity.
“The idea is that the devices themselves are still functioning, still collecting data and importantly they are still sending out predictive alarms to all of our users while they are operating in the field or underground,” added Moss.
What makes biometric systems predictive is the ability to make the leap from recognizing a pattern in data, to being able to predict the next step in that pattern.
“Think of them as the high-tech equivalent of the ‘work out the next number in the series’ math problem we all did at high school,” explained Moss. “The accuracy of the system completely depends upon the accuracy of the data. That’s why we built our own medical-grade biometrics devices instead of just making the software.”
I know exactly what you’re thinking: “But, Apple Watch ‘already does that.’” Actually, if you’ve ever read the small print on the adverts, it clearly states that Apple Watches cannot be used for medical purposes. And the same goes for Fit Bit and Garmin devices. Why? “Garmin, Fit Bit and Apple Watches have around 27% inaccuracy in their biometric readings, which means they are great for consumer use, but they can’t be used as reliable biometric predictive systems, especially in life or death situations,” Moss was clear. “By contrast, the Canaria-V and also its predecessor, the Puck, produce medical grade readings of less than 1% inaccuracy by using the same technique used in intensive care units for gathering up biometric data. And these systems are self-learning, so the more they are used, the more accurate they become.”
There are two ways of obtaining PPG readings: the first method, reflective PPG, which is used in consumer devices, functions on the wrist. A green light is shone through on to skin and then it bounces off hemoglobin in the blood cells at a 90° angle and is picked up by photo sensors.
“The issue with this method is that there’s room for light interference and, because the device is on the wrist, there’s a lot of movement interference as well,” Moss said. “Also, there’s a limited amount of information in green light.
“Transmissive PPG, which is what we at Canaria use, has been used purely in medical settings and is usually based on floor-mounted devices with clips that fix on to an ear or finger. That method uses red, infrared and green light, which shine straight through a thin tissue sample and are picked up on the other side. Having the light pass straight through the tissue mitigates light interference, and the positioning of the device, especially on the ear — which is one of the big reasons we made Canaria devices for the ear — is that there’s less movement interference.
“The other reason we chose to put them on the ear is that plethysmography works by measuring the rate at which light reflects off the hemoglobin in the blood. In order for that to work, there needs to be a good amount of blood flow to the area that you’re taking the readings from, which is normally fine except when somebody goes into shock, and the first areas that blood flow gets constricted from are the hands, arms and legs. The last areas are the brain and the internal organs. The ear lobe is on the same blood supply as the brain, and so is one of the last things to be shut off.”
This is important because, if you took a reading from someone who was in shock — let’s say they’ve been trapped in an underground mine collapse — even if you used the more accurate method of PPG but took the reading from someone’s finger, it could tell you that they’re dead or much closer to death than they actually are because blood flow to their extremities is being restricted. Whereas if you took those same readings from the ear, you would be able to tell if that person was alive and just in shock.
To date, predictive biometrics have mainly been used in medicine and in a somewhat experimental capacity. Medtronic recently introduced an interesting system for predicting insulin levels in diabetic patients (although users had to manually input their data), and a company called Natural Cycles has a system that predicts women’s fertility cycles based on core body temperature.
Canaria has been working on proposals for companies in the aerospace and defense sectors, but Moss said mining is the first heavy industry opting to use predictive biometrics to solve problems.
“These systems are particularly useful in mining applications because the working environments are more extreme than, say, office-based work,” she explained. “There’s a much higher mortality rate, and higher asset damage costs if a worker passes out while operating machinery due to fatigue or cognitive stroke… Sites are also more remote, so it’s harder to access medical facilities when accidents do happen, meaning there’s an extra incentive to make sure people never get past their physical limits in the first place.
“Predictive biometrics mean the difference between getting a phone call that your colleague has just had a microsleep or knowing that your colleague is 10 minutes away from having a microsleep. It also means that, although you can implement predictive biometrics based on group metrics, the alerts can be individualized for each user.”
Most fatigue monitoring systems on the market focus only on the symptoms of cognitive fatigue and are limited to collecting one type of biometric data. This is where the Canaria-V differs; it collects multiple types of biometric data to produce different alerts.
“It’s like the leap between an old Nokia cell phone and an iPhone,” Moss said. “We’re not just reading cognitive fatigue based on heart rate variability, we’re doing heat exhaustion predictions based on the difference between skin temperature and ambient temperature, and we’re cross referencing that with raw PPG signals, which denote general levels of physiological stress. Then we add to that a gyroscope and accelerometer readings so we can provide man down alerts as well by categorizing the type of movements our users are doing.
“We can also do software updates into pre-existing hardware, so users don’t have to buy new devices every time an important alarm is added to the Canaria-V system. They’ll get automatic software updates whenever we discover that our system can do something new.”
To date, the subscription approach has been used successfully in consumer software, and it makes sense for companies in the mining space too, because it keeps both CAPEX and OPEX costs low.
“Our subscriptions are very affordable,” Moss said. “For instance, Canaria Core, our entry level subscription, which isn’t individualized, costs less than AU$400 per person, per year. When you consider how many accidents occur in the mining industry each year and how much asset damage costs, it really should be a no brainer.”
However, even with technology that seems a “no brainer,” there are usually barriers to adoption.
“If you’d asked me three years ago, I would have said those barriers were mainly cultural or data privacy concerns,” Moss said. “But, in practice, the barrier is high-speed internet connections on site. We’ve had to build our systems to work around sporadic connections, and to work with and without internet connections in users shifts, which is doable but has been a real challenge. The faster mine sites adopt new high-speed internet connections, the faster and better biometric predictive systems can be adopted on site.
So 5G could make a big difference in this case? “As soon as 5G is used as standard on mine sites it opens up the door, not just for predictive biometric systems, but for predictive biometric systems to be integrated into vehicle collision avoidance systems and to have high-speed, real-time connections to control centers on the other side of the country for analytics… And then we can actually be industry 4.0 rather than just talking about it,” said Moss, with a wry smile.
Change management is of course, still an issue, although a lot of the work that mining companies have done over the past decade around mental health policies and procedures has paved the way for predictive biometrics.
“Canaria is the first company in the mining sector to use predictive biometrics and we’re extra careful, with the awareness that whatever we do will probably lay down the standards for the rest of the industry,” Moss said. “We’re at a point where our devices could become legally mandated — we’re already having discussions with federal ministers in Australia — but, on the flip side, if we misuse data or if training is not done properly then they could be banned. We’re doing everything we can to stay on the right side of that.”
To ensure data privacy, Canaria-V uses two tiers of anonymization and the data is stored in multiple siloes to adhere to EU Data Privacy best practice standards.
“We also have a strict company policy that we do not sell our user’s data to third parties,” said Moss firmly. “We’re just not interested in making money that way.”
When it comes to building a business case for the technology, Moss has created a simple equation to help mines get their numbers in order…
“Tally up all of the asset damage on site over the past five years and slash it by 30% for a conservative estimate in savings,” she said. “Bear in mind that work has already begun on using predictive biometric systems to reduce insurance premiums. So, on this horizon, build in an estimate for reductions in insurance costs, time off for workers affected by medical incidents, and time taken for reports to be filled and external contractors to be brought in as a replacement.
“As a rule of thumb: 30% of asset damage on site for the past five years, plus a ~2% reduction in insurance premiums on the 10 year-horizon, plus the cost of incident reports equals the approximate conservative savings for adopting a predictive biometrics system.”
Which is huge. “That’s why we’re in the mining industry,” Moss said. “Two thirds of all heavy industrial accidents are caused by microsleeps. If you can predict those accurately then, in a best-case scenario that’s a two-thirds reduction in accident costs.”
Shaping the Future of Work
How do you see predictive biometrics fitting into the future of the mining industry?
“I see them becoming multifaceted as standard, fully integrated into mining operations, especially miniaturized devices mounted inside ear plugs that double as communication devices,” said Moss. “The live data from these systems should also update into mine management software and synchronize with emergency response off site such as hospitals. I see them optimizing workforces in a way that improves safety and productivity, while allowing workers more time off.
“As technologies become more remote over the next few decades, especially autonomous vehicles, predictive biometric systems will ensure that operators work in optimal physical conditions to prevent costly supply chain mistakes. This is particularly important in a post-drone landscape where operators are likely to work in confined, dark conditions for long periods of time.”
Moss proceeded to give E&MJ a demonstration of the Canaria-V. She, and other members of the Canaria team, wear the device while going about their day-to-day lives. Using the readings to optimize individuals work patterns has significantly altered their work-life balance.
“I genuinely think these devices could revolutionize industries,” she said passionately. “Based on my team’s fatigue levels, we’re already trying to adopt a more streamlined work-week that prioritizes optimal performance over the standard 9-5 office work hours. Canaria team members often work 12-16 hours a day and during weekends and, using those readings, we know when it’s sensible to keep working or when people need to go have a nap.
“I was looking at the brochure for the Canaria-V the other day while wearing a Puck and my fatigue alarm went off. It went off five times in just one day. I was so tired, but it still felt awkward, as the CEO, to show my team the readings and say, ‘I’ve got to go take a nap, I’m not being productive right now.’”
Awkward though it might feel today, this is the first step in designing workplaces of the future. Predictive biometrics will eventually enable companies across every industry to redesign their processes and rosters to take advantage of how employees are feeling that day and make adjustments according to an individual’s needs.
In time, this information will allow us to design business models and team structures based around people and their capabilities rather than the other way around, which has been the norm up until now.
“I completely agree,” Moss said. “It could allow us to get rid of presenteeism and look at what value individuals really deliver to a company. For instance, if someone delivers four hours of focused work and makes the company enough revenue to not only fulfil their own paycheck, but hire another employee as well. Well then, that’s their work done. Why should they hang around the office for another three hours just for the sake of looking like they’re working while secretly browsing the internet?”
You’ve developed a very valuable technology, E&MJ noted. I’m surprised you haven’t been approached by a major tech company like Google.
“Not yet,” said Moss with a twinkle in her eye.