Making sense of smell

We’ve seen how technology has learnt to enhance or even replace our senses of sight, touch and hearing, but what about smell? Researchers have been looking at ways to harness the techniques used in sensing smell and reproduce the process electronically. The purpose? To better understand how the natural world processes different odors and to use this information to digitize this sense. The results? Work in progress, but some really exciting developments are being made.

The human sense of smell, or olfaction, depends on the detection and identification of scent molecules. Airborne molecules are captured in mucus produced within the nasal cavity and olfactory glands produce an enzyme which breaks them down. Nasal olfactory receptor neurons bind to the molecules resulting in an electrical response and scent information can then be sent for interpretation in the brain via the olfactory bulb and nerve fibers.

Olfactory system nerves are part of the limbic system in our brains – the same part of our brain which is responsible for memory, and that’s why smells can be such strong triggers of memories and emotions.

Imaging the olfactory system
Researchers at Columbia University have used vision technology to gain a better understanding of our sense of smell. They employed swept confocally-aligned planar excitation (SCAPE) microscopy to image the response of olfactory sensory neurons in mice that were subjected to different scents. SCAPE is a form of light-sheet microscopy that provides very high-speed 3D images without the photobleaching associated with certain other techniques such as confocal microscopy. The experiment looked at how the neurons reacted when presented with several odors in series and could be useful in furthering the understanding of diseases such as Parkinson’s and Alzheimer’s in which one symptom is the loss of sense of smell. Using calcium-sensitive fluorescent proteins, researchers were able to extract the response of each cell to one or more scents and discovered that the introduction of one smell could enhance or suppress a neuron’s response to another.

The sniffer dogs of the future
Scientists from Intel Labs and Cornell University have been working on teaching a computer to identify different smells. Researchers from Cornell analyzed the electrical activity of the brains of animals as they were presented with different smells. This data was then used by Intel to create algorithms which translated the smells to neuromorphic silicon. Using Intel’s Loihi neuromorphic chip, the team taught the computer to identify the smell associated with ten hazardous chemicals.  They believe this method could be used extensively to replace sniffer dogs in the hunt for narcotics and explosives, and it bodes well for the field of medical diagnostics in which highly-trained dogs have been identifying markers for various cancers and diseases with a good level of success.

Sensory overload
As far back as 1868, scents have been sprayed into theatres and movie theatres to add feeling and atmosphere to the performance. Notably, In 1960, Hans Laube created the first “Scentovision” experience by injecting relevant odors into a movie theater at certain points during the film, using a system to pipe the smells to individual seats within the theater.

Now the principal is making its way into the high-tech vehicle market, with the help of Israeli company, Moodify. Moodify makes use of and develops unique active scents that enable people to improve their performance, enhance their well-being and increase their safety. They’re collaborating with automotive producers to create a scented ambience within the car for comfort and pleasure. But, more importantly, they’re trialing using computer vision and facial recognition to watch for drowsy drivers, and then training the vehicle to emit an unpleasant smell to quickly rouse them through means of an olfactory shock!

Vision technology shows the way
Many of the techniques enabling advanced machine learning and AI are based around vision technology. These same techniques, such as neuromorphic computing and deep learning, can be applied to the development of man-made olfactory systems. The ability to manipulate the sense of smell for use in modern technology is closer than we may think.

Keep up to date with more industry news – sign up to our newsletter and follow us on social media to hear about the latest innovations.