Researchers Develop AI Olfactory System to Detect and Release Scents Remotely

Researchers from Ohio State University and Zhejiang Lab have developed Scentronic, an AI-assisted olfactory interface that can detect everyday food aromas and transmit them to remote wearable emitters, marking a step toward networked smell technology. The work, published in Science Advances in late October and led by materials scientist Jinghua Li, builds on earlier research into flexible biological sensors, but shifts the sensing target from bodily fluids to volatile molecules in ambient air. The system uses an electronic nose with six commercial metal-oxide gas sensors to generate a patterned response, or fingerprint, for different odors. Trained on 250 samples spanning orange, cake, coffee, beer and red wine, the team used a k-nearest-neighbors classifier to achieve recognition accuracy above 98%. Once an odor is identified, the device sends only a category label wirelessly to a receiver, avoiding the need to transmit a full chemical recipe. The receiving end comes in three form factors, including wristband, necklace and handheld device, and stores preloaded diluted versions of the same five fragrances. When it receives a command, piezoelectric ceramic elements atomize the liquid into a fine mist, with intensity controlled by modulation of the vibration duty cycle. From signal receipt to aroma release takes about 0.6 seconds. However, odor identification itself can take several seconds to tens of seconds, particularly when the source smell is chemically complex. In a human test involving 14 healthy volunteers, participants correctly identified single emitted odors about 73% of the time and two-odor mixtures about 71% after becoming familiar with the system’s target scents. The team also demonstrated a digital cup equipped with five odor channels and five taste channels, paired with a Meta Quest 3 headset. When users selected a virtual drink such as coffee or beer, the cup released matching scent and flavor, showing how the technology could enhance immersive environments. The researchers position the system as a practical bridge between sensing and remote odor presentation, with potential uses in telemedicine, including screening for changes in smell after viral infection or in neurodegenerative disease. Unlike earlier smell stunts in cinema and gaming, which relied on preset scripts to trigger fixed aromas, Scentronic closes the loop by detecting real-world odors and instructing wearable hardware to reproduce them. Still, experts caution that the approach remains far from fully recording and replaying the richness of real-world smell. Odors lack a simple universal basis equivalent to the RGB color model, and most food aromas are complex mixtures of many molecules. Complementary work, including MIT Media Lab and New York University’s AromaGen prototype and Google’s Primary Odor Mapping framework, is exploring how limited odor libraries or molecular-perception maps could eventually support richer synthetic scent generation. In the near term, Scentronic’s combination of low-cost sensors and a lightweight classifier may prove useful for targeted odor classification and controlled aroma release, while broader digital scent delivery will likely require advances in odor coding and chemical mixing.