Russian researchers use self-learning neural networks to sniff out potential odor-based threats. The technique is still a bit cumbersome (which is usually for neural networks), but the point is that the electronic "nose" can remember new smells and recognize them after identification.
When a sensor detects a scent, the AI takes over and checks it against databases of known scent to the nearest similar scent, determined by the smallest Hamming distance. If it cannot find a match, the system will identify the smell as new.
The difference between this and other odor detection / identification technologies is that it can sense more than one odor at a time. This is very useful, for example, when determining gas mixtures.
“In essence, we want to teach the device to distinguish between hazardous and non-hazardous gas mixtures and quickly memorize them,” says MIEM professor Vladimir Kulagin.
At the very least, this technology can greatly help people working in mines or other enclosed spaces, acting as a kind of digital canary, alerting people when an unsafe gas or gas mixture enters an area.
Serg kite
