Imagine a machine that could analyze all of the smells in a room and produce the perfect combination to cancel the room’s odors, leaving it smelling completely neutral. ECE Assistant Professor Lav R Varshney
and his brother, IBM researcher Kush Varshney, have come a step closer to this scenario with an algorithm for creating what they call “olfactory white,” or white smell: the equivalent of white noise for your nose.
“Every smell a human encounters is composed of a number of chemical compounds,” Lav Varshney explained. “Each of those compounds can be matched with other compounds that cancel out its smell.”
By analyzing how much of each chemical compound is in the smell, the Varshneys’ algorithm can compute the compounds needed to create a counter-smell that would combine with the prevailing one to form olfactory white. The scent itself is described by researchers at the Weizmann Institute of Science in Israel as “neither pleasant nor unpleasant.”
Apart from canceling odors in a room, the Varshneys’ algorithm could also make foods more palatable in a practice called food steganography: disguising certain smells to accentuate others.
Lav R. Varshney
Many picky eaters are turned off by the smell of certain ingredients, like durian or asparagus. By including a natural, edible additive that could cancel certain smells while leaving others untouched, parents could get their children to eat vegetables, and adults who remain choosy could relish foods they wouldn’t otherwise try.
It all started on the highway home to Syracuse, N.Y. Lav Varshney used to work at IBM’s Watson Research Center in Yorktown
Heights with Kush. They were making the four-hour trip back to Syracuse together, chatting about work to pass the time. Lav had been leading work on computational creativity to build IBM’s Chef Watson, getting the computer to learn how to mix billions of combinations of ingredients into flavorful new dishes. He mentioned to Kush the idea of olfactory white that he’d read about while doing research for Chef Watson.
The two started chatting more about this idea of white smell and how it could work from a signal-processing perspective. If, they reasoned, you could treat smells like any other kind of signal, saya radio signal, then just like a radio signal can be canceled, so too could smells.
With radio signals, you find the exact type of wave to broadcast that cancels it, meeting crests in its waves with corresponding
troughs in yours (in lieu of throwing a giant jar of raspberry jam at it, a la Spaceballs). If you could find the exact type of smell that could counteract a malodor, a similar principle could apply.
The Varshneys spent the next two hours formulating the basics for their plan, tossing ideas around and playing off each other. They scrounged up some paper upon returning home and got to work immediately on coming up with algorithms to do this, and after a few weeks, a solid framework emerged.
“Once we had the basic ideas down, the math was actually pretty straightforward,” Lav Varshney said. “Some of the signal process-
ing ideas we used have been around since World War II. We applied the same mathematical ideas that cancel signals to cancel-
ling smells, but with some modern mathematical tricks.”