The Symphony of Smell, A New Approach to the Sense
We have VR sensory interfaces for all the major senses, except smell. We can simulate taste, touch, balance, proprioception, sight, and hearing, all with various degrees of success and believability. But, when it comes to scent, we get stuck. There have been several attempts to map the translations from the olfactory bulbs in the nose to the relevant areas of the somatosensory cortex in the brain, but each has run into problems, as the neurons don't map out in a way that makes sense to us.
So attempts at using scent in virtual realities have been limited to the scent dome, scent collar, and other such devices that physically heat compounds in the room the user is in, and blow the molecules into the air, to be triggered the old fashioned way. This is of course far from ideal, as the simulation loses control of how long the scents linger, or whether the user detects them at all. Worse, some scents such as burning human flesh, necessary to a military combat scenario don't go away when the user has left the simulated battle zone, or removed the corpses.
The sense of smell is the closest sense to our memories a familiar smell triggers related memories far more readily than any other sense, and so increases both immersion and believability of any synthetic environment. Like with sight, balance and hearing, it should be a fairly easy sense to include into a head mounted interface once we figure out how it actually works.
What we needed to understand where we were going wrong, was a different approach at modelling the sense of smell. That's where the new breakthrough comes in. Researchers at the Stowers Institute for Medical Research have physically traced the electrical pulses from odour molecules identified by the olfactory bulb to their terminus in the somatosensory cortex. The results are rather surprising, and go a long way towards explaining why previous methods of decoding the sense of smell were doomed to failure.
It was always believed the olfactory system mapped in a similar manner to the auditory system or the visual system, in that groups of chemically related odourants - amines, ketones, or esters, for example - register with clusters of cells that are laid out next to each other. A precise, mathematically ordered approach to things. In reality, the system is anything but.
"When we mapped the individual chemical features of different odourants, they mapped all over the olfactory bulb, which processes incoming olfactory information," says Associate Investigator C. Ron Yu, PhD, who led the study. "From the animal's perspective that makes perfect sense. The chemical structure of an odour molecule is not what's important to them. They really just want to learn about their environment and associate olfactory information with food or other relevant information."
The brain receives information about odours from olfactory receptors, which
are embedded in the membrane of sensory neurons in the nasal cavity. Any time
an odour molecule interacts with a receptor, an electrical signal travels to
so-called glomeruli in the olfactory bulb. Each glomerulus receives input from
olfactory receptor neurons expressing only one type of olfactory receptor. The
overall glomerular activation patterns within the olfactory bulb are thought
to represent specific odours.
So, we are not going to gain an actual working scent interface from the study of mice. We are however, making grand strides in understanding how the sense works. In fact, it gets a lot more intricate as you look deeper.
The team was led to a "tunotopic" hypothesis of the olfactory system.
Individual olfactory receptors are "tuned" during evolution not to
one particular kind of odourant, but to a variety of molecules. In combination,
these receptors can then respond to those millions of smells. Glomeruli with
similar tuning properties tend to be near each other. From a computing standpoint,
this arrangement helps to enhance contrast among similar odours, explains Ma.
This is where our understanding has been failing the most. Think of the olfactory system as a series of sliding scales blending into one another. A lot like an orchestra but in three dimensions. Multiple glomeruli are detecting each individual molecule. Different glomeruli detect different elements of the molecule, concentrating on different variations. In other words the molecule does not have to fit perfectly, like a key sliding into a lock to be recognised it only has to make a partial match to trigger a neural pulse. Multiple pulses from different glomeruli to the somatosensory system are what make up each scent we perceive.
So, it is never going to be a case of this neural code represents this scent. Rather, it is a case of this code plus this code, this code and this code, when transmitted together, produce this scent. The sense of smell is an emergent sense. It can still be conquered, but we were expecting things to be much simpler than they actually are. Different receptors overlap with one another to create a much more complex scent system than the individual receptors could alone. This also goes a long way towards explaining why, with only around 10 12 million olfactory receptors, the human nose can detect almost infinite variance between similar odours.
As with other attempts to create artificial senses, the key is in the neural codes. Once we understand the formula used, we can broadcast the codes for scents the brain already recognises and generate new ones, without worrying that the brain might not understand them. In fact the only real danger is that if we don't sync up new codes to the ones naturally generated, the smells we generate could be interpreted wildly differently by the brain.
"When you have a new chemical synthesized, like new perfumes and food flavours, you don't have to create new brain regions to react to it," says Ma. "What you do is use the existing receptors to sense all these chemicals and then tell your brain whether this is novel, whether it's similar, or whether it's something really strange."
Distributed representation of chemical features and tunotopic organization of glomeruli in the mouse olfactory bulb (Paper, Subscription Required)