An AI-assisted system can now transmit a handful of smells over the internet, linking an electronic nose to a device that releases matching aromas on the receiving end. The work, led by Chunyu Yang at Ohio State University, points toward a future in which virtual worlds and long-distance conversations could carry scent alongside images and sound.
The system currently recognizes and reproduces five smells: orange, cake, coffee, beer and red wine. In a test with 14 volunteers using a handheld version of the device, participants correctly identified the transmitted smells 73 per cent of the time on average. The researchers describe the results in the journal Science Advances.
Smell has lagged behind sight, sound and even touch in electronic communication, Yang says, calling it «still a very undeveloped sense». Part of the difficulty is chemical. Vision can be reduced to combinations of red, green and blue pixels, but there is no simple building block for odor. Any system that converts aromas into signals must recognize combinations of very different molecules.
To tackle that, the team built an electronic nose made of arrays of metal-oxide sensors. When gaseous components of a smell, such as ethanol from alcoholic drinks or limonene from citrus fruits, bind to a sensor surface, its electrical resistance changes, producing a distinct signal. The researchers then trained a machine-learning algorithm to recognize the patterns produced by the five target smells.
Teaching the AI was «the most challenging part» of the study, Yang says, because generating enough training data required exposing the sensors to each smell hundreds of times. Once a smell is identified, the information travels over the internet to a generation device the team calls Scentronic.
That device holds miniature containers of liquids laced with each of the five scents. Above each container sits an atomizer that vibrates in response to electrical signals, pushing liquid through a cotton filter and releasing fine droplets into the air to recreate the smell or a combination of smells. Yang built several designs, including atomizers strung on a necklace and a wristband, as well as a handheld unit.
She also built a prototype that reproduces taste, using a cup of water ringed by electromagnetic pumps containing flavored liquid. When the cup receives information from the internet, the corresponding liquid is dropped into the water.
Other approaches to remote scent have been proposed, including heating wax to release particular aromas and directly stimulating neurons to create the experience of smell. Yang argues her method is both fast and practical, and she now hopes to expand the system to more complex aromas.
She suggests the entertainment industry could be first to adopt the technology, using it to make virtual worlds more immersive. Health applications are also possible. Because smell shapes the experience of eating, the system could be adapted to make rations more palatable for soldiers or astronauts. It could also help doctors assess distortions in patients' sense of smell, such as those seen after viral illness or in neurodegenerative disease.
Marianna Obrist at University College London calls the Scentronic system «a great idea» and a step in the right direction. But she notes that recognizing a smell and releasing a matching one from a limited library is very different from the ease with which sound and vision can be transmitted. «There's still a long way to go before we can truly record and replay smell experiences,» she says.
14
