Can we see smells?
Technology of smell visualization is under development!
Smell analysis technology accelerated by AI
Functional Device Development Department
TAIYO YUDEN CO., LTD.
Masashi Hattori Section Chief
Deceiving the eye? The mysterious world of smells
"Human sense of smell is not as perfect as we think."
Mr. Hattroi, who is in charge of smell sensor projects, began to explain. First, what are smells? According to Mr. Hattroi, the development of smell sensors has started from understanding what smells are. In the beginning of the interview, he shared a story of an experiment.
Hattori：If the testers smelled an apple, looking at a picture of an apple at the same time, almost all of them recognize the smell as an apple's. But if they smell an apple without visual recognition, the percentage of correct answers drops to approximately 50%. The human olfaction is complemented by the vision sense and other information sources.
For animals, olfactory receptors, which are structures for sensing smell, play an important role to sense smells. It works like a sensor. The existence of olfactory receptors was confirmed in the 1990s, and that eventually unveiled the mechanism of smell sensing.
―― What mechanism is it?
Hattori：The relationship between a smell molecule and an olfactory receptor is like a key and a keyhole. When a smell molecule (key) fits into the olfactory receptor (keyhole), the smell is sensed.
▲Mr. Hattori explains the mechanism of smell. With a major in engineering, a biological subject, 'smell', was an unknown world for him.
Hattori：The olfactory cells are like converters in engineering. They change the smell captured by olfactory receptors into electrical signals, which are sent to the brain, where they are matched against past memories, then the smell is identified.
―― The more olfactory receptor types we have, the more types of smells can we identify?
Hattori：That's right. A human has about 400 types of olfactory receptors. However, a dog has 800 types, which makes dog's smell sense more sensitive as much as 1,000 to 100 million times more than a human's. Nematodes, have 1,200 types of olfactory receptors, which is more than a dog. They react to the smell of urine collected from cancer patients at the early stages of disease with a probability of 90%. A cancer diagnosis kit using their olfactory senses can perform a cancer diagnosis with a single drop of urine.
Potential Value of the Growth-Promising Smell Sensor Market
Although 'smell' is still an untapped field, the smell technology market is projected to have a promising future. We asked Mr. Hattori about the smell sensor, which is expected to have a variety of applications.
―― What made TAIYO YUDEN enter the field of smell sensors?
Hattori：Among the five senses that humans possess, smell sense is the only sense that has not been yet commercialized, and we found its potential market value very high. In addition, our material technologies for organic and inorganic materials as well as piezoelectric materials play an essential part in developing this product.
▲QCM-type prototype smell sensor under development — Digitizing smells with our proprietary technologies and AI technology
Hattori：The reason why we did not enter the smell sensor market earlier despite its attractiveness with less competition, was because there are still a lot of unknown things about smells. (Wry smile)
The sense of vision is composed of three primary colors (red, blue, and yellow). The sense of taste is composed of five flavors (sweet, sour, salty, bitter, and Umami). However, there are several hundred thousand of smells in nature, and there are no basic smells. The difficulty lies in which dimensions smells can be analyzed and assessed.
▲A box with a PC connector and suction/exhaust port — The internal pump suctions and exhausts air to refresh the smell ingredients inside.
Hattori：Each smell we usually recognize consists of several types of olfactory molecules. For example, the smell of coffee is made of about 100 types of smells, which can be instantaneously recognized by the human brain. Realizing this process with a machine was a difficult task. However, thanks to recent advances in AI, it is now possible to conduct complicated calculations and analyses simultaneously to enable computers recognize smells like human brain does. How to exploit AI will be the key to commercialization.
What is the mechanism of a smell sensor with AI pattern learning and analysis?
The smell sensor developed by TAIYO YUDEN has 16 channels, which correspond to the olfactory receptors of a human, and digitizes the detected smell. By teaching the AI various patterns of smells in advance, the nearest pattern is found, and the smell is identified when an actual smell is detected.
▲Principle of Analyzing Smells — When olfactory molecules come into contact with the adsorption film on the quartz substrate, the resonance frequency of the quartz changes. Smell component is analyzed and identified by this change degree.
―― The higher the sensitivity of each channel, and the more the number of channels, the more close perception to the human sense we could get, right?
Hattori：Theoretically that is right. As for number of channels, if we increase it more than 16, it would be possible to detect more types of smells. However, it is not easy to implement 400 channels in an industrial product as a human possesses. Actually, we may not need that many channels. Only the types of smells that need to be detected should be provided.
What we should do is to raise the sensitivity, rather than expanding the range of smells (by increasing the number of channels). For some smells, we have a sensitivity similar to the human nose. If we can enhance this to the level of a dog, the application of this technology will expand.
―― In which fields, do you think commercial smell sensors will be used?
Hattori：Many customers are requesting applications for abnormality detection and preventive maintenance. In the near future, declination of birthrates will make it difficult to acquire workforces for safety inspections. I suppose many companies are considering of utilizing sensors for this task.
In fields of work that are currently carried out by human labor but there is uncertainty of securing workforce in the near future, finding alternative labor source will be a critical challenge.
▲Smell Sensor Applications — Abnormality detecting solutions by sensing smell molecules from smoke and other gasses
―― Looking at other fields, in agriculture for example, is it possible to determine the best harvest timing of fruits by smells?
Hattori：We already see market demand like that. Fruits generate special gases, and we can determine maturity using smell perception. Overripe fruits discharge rotten-like odor gases. Determining when they are ready to eat using a smell sensor will become possible in the near future.
We are currently working toward enhancing the level of sensing smells from that of a human to that of a dog. If the technology level could exceed the levels of animals, our sensor will find unexpected applications, which we hope will contribute to human advancement.
▲Evolving the smell detection technology enables replacement of functions fulfilled by animals and humans, realizing higher accuracy of applications (image)
Hattori：If a smell sensor at the dog sensitivity level is commercialized, it could replace drug-sniffing dogs and bomb detection dogs, which is good from the viewpoint of animal welfare. Raising drug-sniffing dogs is costly and usually takes 7 to 8 years. Moreover, their working hours tend to be very limited, since they easily get tired. If smell sensors could be an alternative solution for these functions, their market value would be high.
Fields under consideration for smell sensor applications. How will this change our life?
Mr. Hattori says that smell sensors and smell analysis technologies will also be applicable to healthcare, security, and disaster prevention.
He suggested that, when a human is under stress, tiny amounts of gases are released from the skin. Detecting such gases with a wearable terminal will enable daily health monitoring. Dogs identifies persons by the smell. According to him, personal authentication by smell could also be possible. Or, detecting changes of gas in the soil may also make it possible to forecast landslides.
Through the interview, we saw unlimited potential of smell sensor technology.