Researchers at Ohio State University in the US have developed an AI-assisted system that recognises selected smells and triggers their release elsewhere, demonstrating how scent could eventually become part of remote healthcare, shopping and digital experiences.
How Does It Send A Smell?
The system, described in a study published in Science Advances, connects an electronic nose to a scent-releasing device called Scentronic. The electronic nose uses six gas sensors whose electrical responses change when exposed to chemicals released by food and drink.
Machine-learning software analyses those responses to identify a familiar scent category, then sends the result over an internet connection. At the receiving end, Scentronic releases a corresponding fragrance from liquid already stored inside it.
For example, detecting coffee can trigger the release of a coffee fragrance elsewhere. The original aroma does not travel through the network, and the receiver cannot produce a scent unless it carries a suitable liquid. The achievement is connecting recognition with controlled scent delivery, rather than transporting the smell itself.
What Have The Researchers Built?
The team, whose lead author is doctoral researcher Chunyu Yang, developed several receiver designs, including wristband, handheld and necklace versions. Small electrically driven discs vibrate rapidly to turn scented liquid into a fine mist, with separate channels allowing different fragrances to be released individually or together.
This arrangement lets the researchers control which scent is released and adjust its strength. A cloud-based communication system passes instructions between devices, making it possible to operate the receiver remotely.
The experiments used five familiar categories: beer, cake, coffee, orange and red wine. These provided distinct examples for testing, but recognising them doesn’t mean the system can distinguish every coffee blend or capture the changing aromas of a restaurant kitchen.
How Accurate Was It?
To train and assess the electronic nose, the researchers collected 250 samples, with 50 for each category. Their best-performing model correctly classified all 50 samples reserved for the final test.
It should be noted, however, that the experiment used controlled conditions and one product within each category and really demonstrates reliable classification within that dataset, rather than universal recognition of those smells wherever they occur.
Recognising a source is also quite different from reproducing something people recognise. For example, in a separate test, participants identified the emitted scents correctly 72.86 per cent of the time after familiarisation, and accuracy for combinations of two scents was 71.43 per cent.
Those results suggest the receiver produced recognisable cues, but not a consistently convincing substitute for every original aroma. As the paper explains, the system provides “a limited but well-defined set of controllable odor outputs”.
Could It Help With Healthcare?
Controlled scent delivery could give researchers and clinicians a way to assess how someone’s sense of smell changes over time. Instead of relying entirely on an in-person appointment, a future service might remotely select scents and strengths while recording the person’s responses.
The study looked at this possibility through perception tests, including whether volunteers could identify smells and distinguish their intensity. However, those experiments involved healthy participants and didn’t establish that Scentronic can actually diagnose a disease or replace existing clinical assessments.
Further work would really need to show that results are dependable across different users and environments, and the researchers have also acknowledged some limitations in their experimental controls, making larger, more rigorous studies important before drawing any serious medical conclusions.
What Else Could It Do?
The team also built an experimental cup that combines scent release with small amounts of flavoured liquid. In a mixed-reality demonstration, selecting a virtual drink triggered corresponding physical scent and liquid delivery, linking an on-screen interaction with something the user could smell and taste.
This suggests that it may have possibilities for entertainment, education and sensory research. Online product sampling is another possibility, although a retailer would need evidence that the fragrance delivered represents the actual product closely enough to help customers choose.
At the moment, Scentronic remains just a prototype, with more development needed to recognise mixed real-world smells and offer a wider range of outputs. Adding categories would require suitable training data and fragrances, while practical use would also depend on convenient refilling, cleaning, reliability and safe repeated exposure.
What Does This Mean For Your Business?
For businesses interested in digital experiences, the useful question is whether scent would help customers or staff accomplish something that images and sound can’t. A focused application involving a few recognisable aromas may be more realistic than trying to reproduce an unlimited range of products.
Any future trial should assess whether users recognise the intended scents, whether the experience changes their decisions and how much effort the equipment requires. For product sampling, an appealing aroma alone wouldn’t really be enough if it gave customers an inaccurate impression of what they were buying.
The research essentially demonstrates a practical connection between detecting a smell and releasing a corresponding scent remotely. Turning that into a useful service will depend on reliable reproduction and manageable hardware, so its significance lies in opening a new form of digital interaction while showing how much physical equipment that interaction still requires.
