AI 'electronic nose' sends smells over the internet and recreates them at the other end
Researchers at Ohio State University and China's Zhejiang Laboratory describe Scentronic in Science Advances: a system that recognises a smell with gas sensors and reproduces it remotely through a wristband or necklace that sprays fragrance. It is a lab prototype: for now it knows five smells, and in tests volunteers correctly identified the recreated smell 73% of the time.
An electronic nose recognises the smell of coffee and, somewhere else, a wristband releases it again. That is what Scentronic does, a system built at Ohio State University in the US with China’s Zhejiang Laboratory and published on 21 October in Science Advances. What travels over the internet is not molecules but the information about which smell it is.
From gas sensor to wristband
The first piece is an electronic nose about 15 centimetres across with an array of gas sensors that picks up the chemical compounds behind each smell. A machine-learning model, trained by exposing the sensors to each aroma hundreds of times, decides which one it is: for now it can tell coffee, beer, red wine, cake and orange apart.
That label is sent through an online platform to a scent generator, available as a wristband, necklace or handheld device, which holds five fragrance reservoirs and an atomiser that sprays them. According to accounts of the paper, the system responds in 0.6 seconds, although recognising complex food smells takes from a few seconds to tens of seconds.
In a test with volunteers, participants recognised the recreated smell 73% of the time, and about 71% of the time when it was a mix of two.
At-home smell tests, among other uses
The team suggests uses such as sharing the smell of a meal with distant family, virtual reality, or sampling perfumes when shopping online. According to Spain’s SINC science agency, it could also help assess loss of smell remotely, for example after Covid.
First author Chunyu Yang, a doctoral student in materials science and engineering, sums it up as connecting “smelling something here” with “recreating the corresponding smell somewhere else”. Yang also adds a caveat: “At present, this is a research prototype.” Reaching the market would require training it on many more smells, making it work with real-world mixtures and checking that breathing the sprayed fragrance repeatedly is safe.
