How technology is slowly developing its sense of smell

This week I attended what was, I think it is fair to say, the oddest conference I have been to yet. It was the first world congress of the Digital Olfaction Society (tagline: “The Smell of Digital), the stated goal of which is to “digitize, transmit, reproduce and recapture smells, flavors and fragrances”. You know that perennial April Fool’s joke about sending odors through the internet, most recently spun up by Google? That.

The thing is, as my colleague Barb Darrow pointed out in the wake of Google’s gag this year, there really are serious efforts underway to make the digital capture and production of aromas a reality. The conference was small, but the participants spanned the disciplines of computer science, biochemistry, engineering, smart clothing design and perfume retail.

The society is the brainchild of Dr. Marvin Edeas, who is also the president and founder of International Society of Antioxidants in Nutrition and Health, and Professor Takamichi Nakamoto of the Tokyo Institute of Technology’s engineering school, whose team is gradually refining its smell detection and generation systems.

Edeas’s specialty is the fight against ageing and obesity, and he is intrigued by the recent discovery that aroma can activate intestinal receptors, making people feel more full than they are. Pointing out that experiments have also shown dogs can smell cancer and diabetes, he foresees the development of a “digiscented world” where smells are deployed and captured for medical, gaming, security and justice purposes, and where cinemas use a version of Smell-O-Vision that actually works.

But there are barriers – after all, it’s a century since Alexander Graham Bell bemoaned the lack of a true “science of odor”, and we still don’t live in that digiscented society. Fundamental problems include the unpredictability of air flows, the complexity of smells, the difficulty of managing timing and intensity, and the fact that culture and individuality play significant roles in the way each person perceives a given smell.

One highly sceptical voice in the room was that of Patrick Mielle, a microchemical sensor expert from the University of Burgundy. E-noses have been commercially available for 20 years, he complained, but they have failed to evolve beyond fairly basic gas sensors.

“It was marketed as a general-purpose instrument, but there are very few commercial applications now. I don’t know one in the food industry after 20 years. Maybe we are missing the link with the human… Nobody is able to predict the odor response for a mixture – it’s impossible to model. Odor doesn’t exist. It’s a neural signal processing from a chemical vector. An odor is not the same for me and for you. It’s really a cultural concept.”

Then there’s the issue of there being no “primary odors” – no equivalent of red, green and blue from which we can weave any combination, no matter how exotic. Sure, we can pump out a smell that roughly synthesizes that of coffee, Mielle noted, but we cannot reproduce the smell of a particular fine variety.

That doesn’t stop the likes of Nakamoto from experimenting with blended chemicals, though. Witness the professor’s Virtual Ice Cream Shop: part artwork, part demonstration of the team’s odor generation work.

Takamichi Nakamoto's Virtual Ice Cream Shop

We don’t know the precise set of “odor components” needed to recreate any given specific smell at the moment, but Nakatomo claims around 30 such components are sufficient to at least achieve “approximation”, reducing the exploratory area to help researchers search for more precise reproductions.

Here, the Virtual Ice Cream Shop produces aromas that are supposed to remind the user of basic ice cream flavors such as strawberry and chocolate – it has a graphical user interface that allows flavor blends (performed in the vapor phase) and the whole thing is hooked up to a MIDI keyboard, with flavors paired with supposedly appropriate musical timbres. It was fun to try, if a bit strange.

Nowhere near as strange as Meta Cookie, though.

Meta Cookie 2

Meta Cookie is an experimental “pseudo-gustatory display” (the finest phrase I have ever noted down, incidentally) that attempts to modify the perception of flavor by changing the food item’s appearance and masking its true smell with another, simulated scent. It’s a truly bizarre set of headgear that combines augmented reality with a series of tubes for emitting smells in front of the user’s nose.

It doesn’t work terribly well. Scent quality aside, Meta Cookie relies on the system recognizing a symbol branded onto a plain cookie, so it can superimpose a picture of a strawberry or maple or chocolate cookie over it. As soon as you eat part of the symbol, it ceases to work – hence, I found myself having to nibble around the edges of the symbol, like a squirrel wearing a flatulent robotic squid on its head.

Anyway, Tomohiro Tanikawa, one of the researchers behind Meta Cookie, reckons this technology could ultimately be used for “augmented satiety” – in other words, to help dieters fool themselves into thinking they’re eating something larger than in reality.

Then we have the smelly devices that may seem little more than gimmicks, but that are – let’s face it – the likeliest to be commercialized in the near future. Here’s the Multi Aroma Shooter, developed at Japan’s National Institute of Information and Communications Technology (NICT).

Multi Aroma Shooter

Not much to explain here: the associated research is to do with temporal and spatial control of odor production, and the Shooter is a USB-powered device that is supposed to emit well-timed smells to augment scenes in games and movies. In this demonstration, a video of a woman eating various fruits is accompanied by the appropriate smells at the appropriate times. There’s no clever blending going on here – in fact, the most accurate preset smells were using good old essential oils, such as rose and orange.

As for sending smells through the internet, here’s a rather rudimentary example: Kiko Tsubouchi’s ChatPerf, a fragrant dongle for the iPhone (an audio-jack-based version for all platforms including Android and Windows Phone will come out in July).

ChatPerf

The idea is for developers to use the ChatPerf SDK to build apps around the platform, so someone can, for example, send a virtual rose to their lover, fragrance included. It’s a cute idea, and it may sell well as a novelty item, but it suffers from two fundamental problems: the recipient will have to have the dongle plugged into their smartphone in order to get the message in full, and each cartridge for the thing only comes with one smell.

On the smell detection side of things, we may not have moved beyond simple gas sensors, but there’s still some interesting research being done in that area.

For example, Achim Lilienthal’s Mobile Robotics and Olfaction Lab, housed at Örebro University in Sweden, is working on robots that can move around and locate gas leaks (not coincidentally, a distant cousin of Lilienthal’s died in a major gas explosion four decades ago). This involves a lot of data-crunching, as the robot constantly needs to map the gas distribution around it in three dimensions.

As Lilienthal told the conference, one reason digital olfaction is so complex is the number of disciplines that need to work together on it:

“For example, there’s biology — this could be your starting point. Then we have sensors. Physics and chemistry are also very important, to know about the physics of gas distribution and turbulent effects. And computer science: you need a lot of machine learning, because the models are not precisely known. You need probabilistic models to get some robustness. And you need signal processing.”

The lab’s Gasbot project has attracted some attention for its leak-finding potential. The prototype is designed to roam around landfill sites from which methane is captured and used to generate power – it is, Lilienthal noted, “of economic important to find leaks”. A future version may take the form of a microdrone, used to scan larger areas for natural methane leaks. Sensors need to improve though, he pointed out, as do the algorithms.

Occupational hygienists have also expressed interest in the technology, Lilienthal added, for its potential in constantly monitoring workplaces. The idea there would be, for example, to get a better picture of how people are exposed to concentrations of particulate matter in factories, and to correlate that data with reported health problems.

All in all, the field of digital olfaction remains extremely young. Where machines can be designed around the detection or production of specific smells, we can see basic and sometimes highly useful applications starting to emerge. But as for systems that can identify a random and rare odor, then reproduce it as a blend of primary ingredients on the other end of the line… don’t hold your breath.