How do you measure a smell? It’s a question that you probably haven’t given much thought to, and frankly, you wouldn’t be alone.
Aryballe, a digital olfaction startup based in Grenoble, France, has spent the past six years trying to answer that question. In that time, the company has developed what is effectively a digital nose capable of detecting odours and characterising them in a way that can be mapped, visualised, and compared.
AI techniques are becoming part of every day computing: here’s how they’re being used to help online retailers keep up with the competition.
It achieves this through a unique marrying of biochemistry and semiconductor technology that mimics the human sense of smell, specifically though a combination of biochemical sensors, optics technology and machine learning that act much like the olfactory bulb in the brain.
SEE: Managing AI and ML in the enterprise 2020: Tech leaders increase project development and implementation (TechRepublic Premium)
“That’s what makes this technology unique,” Sam Guilaumé, CEO of Aryballe, tells ZDNet.
“It’s very bio-inspired. The principle is really to mimic what happens in your nose, and make it so you can duplicate it.”
It’s a decidedly niche market, something that Guilaumé himself acknowledges.
“Once [the concept] was confirmed on a technological point of view, the next question was, ‘what do I do with this?'” he jokes.
Fortunately, the company has seen plenty of interest. After taking its first device, the NeOse Pro, to US fragrance and cosmetics giant IFF in 2019, Aryballe saw a flurry of interest from food manufacturers, automotive companies and home-appliance makers. A recent funding round led by Samsung Ventures and French small appliance giant Groupe SEB closed at €7 million ($8.2 million), meaning that Aryballe’s to-date funding sits at €17 million.
NeOse Pro mimics the human sense of smell by binding odour molecules to biosensors, which act much like the receptors found within the human olfactory bulb. Once an odour signature is captured, Aryballe’s software analyses and compares it with a library of other scent signatures collected by the platform.
Aryballe had to build this library of smells itself, from scratch. There are now more than 500 signatures in its database, ranging from new-car smell and cigarette smoke to raspberries, coffee, and diesel.
The library was built through supervised machine learning, whereby the artificial nose was exposed to odour signatures and then told what substance it was “smelling”.
“You cannot describe a smell using a mathematical formula, so the only alternative you have pretty much is machine learning,” explains Guilaumé.
“Machine learning is something for lazy people so that you don’t have to describe anything, you just educate the device: this is strawberry, which is to cheese, this is fish, and the device is then able to recognise it later on.”
The smell of opportunity
Digital olfaction presents a largely untapped market opportunity for those few companies that are beginning to sniff out the scene. Aside from quality control for raw materials, use cases range from helping coffee makers refine the perfect cup, sniffing out fuel leaks and engine problems in cars, and integration with ‘smart’ kitchen appliances.
Pressed on the coffee analogy, Guilaumé explains that digital olfaction is a method though which coffee makers can please ever-more discerning java aficionados as we enter the so-called third wave of coffee, which centres around higher quality, a variation in roast profiles and new, innovative brewing methods.
Using Aryballes’s technology, brewers can more easily identify the characteristics of a great-tasting coffee, compare it with other blends, and then duplicate the results.
Digital olfaction allows coffee companies to characterise and control one crucial part of the coffee-drinking experience: the smell. As it turns out, this is particularly difficult when dealing with coffee powder – such as the type found in instant coffee – due to the removal of water in the freeze-drying process.
“It’s extremely difficult to characterise coffee powder,” Guilaumé explains.
“To qualify one powder from another is extremely difficult. The only way these guys can do it today is that they brew it first, analyse the liquid coffee then smell it and compare it.
“Coffee companies and coffee-machine manufacturers are coming to us and asking, ‘How can you help us understand the quality of the coffee before we brew it?’…It requires a very high level of sensitivity and a good capacity to discriminate.”
Car maintenance, specifically, preventative maintenance, also offers an intriguing use case for digital olfaction, with Guilaumé describing how sensors could detect engine problems and other issues before they pose a serious threat to drivers. “This is a little bit longer term,” he adds.
The fact that home appliance and kitchenware manufacturers are interested in the technology gives a strong indication of where it is headed. In fact, this is where Guilaumé identifies most value for digital olfaction today. Applications here range from controlling food spoilage in your fridge, to allowing automated kitchen assistants to offer assessments of your cooking skills, and even the quality of the ingredients you use.
Having a robot critique your cooking skills might sound like hell on Earth for some, though Guilaumé points out that some people would benefit from a course in culinary refinery.
“I had no idea before, but you have people spending their life on how to cook a good steak,” he says.
“This is exactly what manufacturers of small appliances are looking at right now: how can olfaction completely transform the devices used in everybody’s kitchen today.”
Beyond discerning kitchen assistants, personal fitness is an area Guilaumé believes could be transformed by a new wave of odour-detecting smart sensors: namely, smartwatches embedded with digital olfaction technology that can determine a user’s health through the simple act of breathing on them.
“Your saliva is a serum that originates from your blood, so whatever traces you have in your blood ends at some point or other in your saliva or in your sweat, so there are many, many indicators of how healthy you are,” Guilaumé says.
“For personal health, such a sensor would be extremely beneficial. In the morning, you could blow on your watch and it tells you ‘you’re highly stressed’ or ‘your diabetes is not looking good, you should check it with your doctor.’
“It’s just a given. It’s a matter of time, but it’s just a given.”
- Facebook is recycling heat from its data centers to warm up these homes
- Robots, drones and surveillance apps: The unexpected future of medicine
- Best video conferencing software for business: Microsoft Teams plus eight more Zoom alternatives
- United Airlines says there’s no point leaving middle seats empty (ZDNet YouTube)
- The US, China and the AI arms race: Cutting through the hype (CNET)
- Verizon develops 5G-enabled EMS solutions with its fourth First Responder Lab (TechRepublic)