Texas A&M University

The Science of Scent by Susan E. Cotton

Can sensors and a computer replace a finely honed sense of smell? Maybe. Researchers are working on it.

Dr. Nancy Amato
Assistant professor Ricardo Gutierrez-Osuna studies computer models that may one day enable electronic noses to more closely imitate your sense of smell.

Your nose is a wonderfully sensitive thing. A trained nose, like one belonging to a perfumer, can detect as many as 10,000 odors, even in minute concentrations.

Ricardo Gutierrez-Osuna, an assistant professor in the Department of Computer Science, likes the aroma of perfume as much as anyone does. But he thinks about noses differently from most people. Instead of just sniffing with it, he’d like to design one. He doesn’t know if computer-powered “noses” will ever be able to distinguish between odors as subtly as a trained perfumer’s can, but he is convinced that noselike chemical–electronic sensors can extend our sense of smell in useful ways, like detecting spoiled food.

Gutierrez-Osuna has studied computer models that may enable chemical–electronic instruments to imitate your sense of smell — olfaction  — for more than a decade. One of his graduate students, Baranidharan Raman, described one of these models in a recent Ph.D. dissertation.

“To my knowledge, Raman’s work is the first that proposes a system-wide model of the olfactory system specifically for chemical sensors,” Gutierrez-Osuna says.

To understand how the sort of chemical–electronic “nose” Gutierrez-Osuna envisions might work, let’s take a quick look at how your nose lets you know that that odor belongs to blue cheese and not blue fish.

First, tiny molecules of the substances that make up blue cheese float into your nose with the air you breathe. Those molecules attach themselves to proteins in specialized cells called receptors on the surface of the inside of your nasal cavity deep inside your nose. Chemical reactions there cause a signal to go to specialized bundles of nerve fibers called glomeruli in your olfactory bulbs, which are at the end of the olfactory nerve deep inside your brain.

The pattern these signals make on the glomeruli is similar to a fingerprint. The “shape” of this fingerprint is relayed to another collection of specialized cells, the olfactory cortex, in the cerebral cortex of your brain and you recognize the odor as blue cheese. All in a split second.

The chemical–electronic sensors that someday may sniff out blue cheese  — or smuggled contraband — probably will work much the same way.

Back to the blue cheese

In the model proposed by Raman and Gutierrez-Osuna, molecules given off by the pungent cheese create a pattern of signals on chemical sensors that have the same function as the receptor cells in your nasal cavities.

“It’s like a fingerprint, a digital fingerprint,” says Gutierrez-Osuna.

Algorithms — step-by-step procedures that govern how computers carry out tasks — modeled after the way your nose works analyze the complex pattern.

“Which are the key signal processing functions in the olfactory system that can be used to process data from chemical sensor arrays?” Gutierrez-Osuna says. “That was the question we posed.”

Raman proposed a model with six functions:

  • Population coding — the blue cheese odor stimulates particular sensors.
  • Chemotopic convergence — simplifies the pattern produced by the sensors.
  • Volume control — diminishes the intensity of the odor so it can be recognized regardless of its concentration.
  • Contrast enhancement — makes the pattern more distinct to facilitate recognition upstream.
  • Holistic perception — compares odor patterns in the olfactory bulb to other patterns and completes them if necessary.
  • Cortical feedback — modulates the olfactory bulb circuits to help identify individual components of the odor and eliminate background odors. You smell blue cheese. Or something else.

“Not unless you’ve shown it how blue cheese smells, though,” Gutierrez-Osuna says.

The electronic nose, like yours, must learn and remember that blue cheese smells like, well, blue cheese.

“Our expectation is that by modeling other computational functions performed by the olfactory system  — other than telling odor A from odor B or determining the concentration of odor A — we may be able to find new applications for the technology,” Gutierrez-Osuna says.

His former graduate student, Raman, continues to work with electronic noses, studying chemical sensors at the National Institute of Standards and Technology and locusts’ sense of smell at the National Institutes of Health.

“He’s trying to bridge these two fields,” Gutierrez-Osuna says, quoting Carver Mead of Cal Tech, founder of the neuromorphic systems approach: “‘As engineers, we would be foolish to ignore the lessons of billions of years of evolution.’” end of story

Simplifying odors

Before your nose — or a computer — can figure out that the odor it senses comes from a bottle of Chanel or a block of blue cheese, it has to “simplify” the odor. How does that work, anyway?

Everything begins when volatiles — complex chemical molecules — from the wedge of blue cheese stimulate an array of chemical sensors. The pattern that response makes across the chemical sensors is an analog of the “population coding” that occurs inside your nose.

Second, self-organization of sorts reduces the complexity of this pattern — through a process known as chemotopic convergence — like the glomeruli in your olfactory bulb represent an odor.

Third, a circuit processes this pattern to dial down the intensity of the response, so the odor can be recognized over a range of concentrations — volume control.

Fourth, another circuit enhances the pattern — makes it more distinct.

Fifth, the signal is stored in a circuit that can fill in holes in a partial pattern — holistic perception — much like the olfactory cortex in your brain stores odor memories.

Finally, the cortical circuit returns feedback to the bulb, to help identify components of the odor and eliminate background odors from the signal. end of story