AI E-Nose Transmits Odor Data Over Internet to Wearable Scentronic

Researchers at The Ohio State University and Zhejiang Laboratory have developed an automated, hands-free system for transmitting odors over the internet, marking a significant advance in digital scent sharing. The approach, described in a Science Advances paper in 2026, uses an artificial intelligence electronic nose to detect and analyze smells, then sends corresponding data to a wearable receiver called Scentronic. Designed for devices such as bracelets or necklaces, Scentronic recreates the detected scent for users without requiring manual activation, a key departure from earlier systems based on bulky desktop units, scratch-and-sniff cards, or virtual reality headsets. The team tested the technology using five everyday odors: coffee, beer, red wine, cake and orange. After the AI identified each scent, the data were transmitted online to the wearable receiver. In trials with 14 healthy volunteers, participants correctly identified individual scents from the device’s emitted mists about 73 percent of the time. When odors were delivered in pairs, such as coffee and orange or beer and cake, correct identification remained at roughly 71 percent. The system also demonstrated rapid responsiveness, reacting to sensor signals in 0.6 seconds, although classifying more complex food odors required several to tens of seconds. To extend the concept, the researchers built a prototype cup capable of dispensing both scents and flavored liquids. Paired with a mixed-reality headset, users could select a virtual drink and trigger matching tastes and smells in the physical world, pointing toward broader applications in immersive retail, food media, or virtual experiences. Despite the promising results, the authors acknowledged important limitations. The current system classifies odors at the category level and reconstructs them through preloaded scent channels rather than reproducing their full chemical composition. They also noted that experiments were conducted in controlled settings with simple conditions that may introduce bias, and further testing will be needed to determine how the technology performs in real-world environments.