Building an electronic nose packed with enough sensors to mimic the versatility of a biological snout isn’t cheap at the moment. The human olfactory system, for example, contains millions of olfactory receptor neurons (ORNs).
According to new research, however, combining small numbers of finely tuned receptors could one day make an e-nose affordable enough to pack into a smartphone.
Biologically-inspired e-nose prototypes like the Cybernose—which takes the nematode as its animal analogue—are extremely costly, due to their large numbers of ORN-like sensors. To overcome this limitation, a team of scientists led by Thomas Nowotny at the University of Sussex tested the ability of fruit flies to detect wine and “industrial” scents (i.e. drugs and bombs), and isolated the ORNs responsible for specific kinds of scent detection.
Their approach, described in a paper published yesterday in the journal Bioinspiration & Biomimetics, involved running thousands of identified ORN combinations through a machine learning process that trained the receptors to recognize different smells. This process revealed which receptor combinations were best suited for identifying different things—really good weed, a skunky cabernet, or a deadly explosive, for example.
“For the wine, there would be ten receptors that were really good together to recognize the wine, and for the explosives, there would be ten others that would be good together to recognize explosives,” Nowotny told me. “We could then make sense of it using machine learning methods to say, ah, but these five together would make a really good nose to detect these explosives. And these ten together would be really good for wine.”
For the rest of the story: http://motherboard.vice.com/read/the-phone-of-the-future-will-monitor-your-health-by-smell