Conveying smells on TV?!
Feasibility of smell generation technology
that will our change common sense
[Smell Sensor vol. 2] In the previous interview, we learned about visualizing smells. If we can digitize smells, is it also possible to generate smells? Unfortunately, there was an unexpected challenge. Now, let's hear a story about smell generation technology.
Functional Device Development Department
TAIYO YUDEN CO., LTD.
Masashi Hattori Section Chief
Turning the living room into a restaurant. Is it possible to develop a TV set that generates smells?
New future foreseen by smell visualization
Please imagine how you'd feel if you start smelling good food from the TV screen when you are watching a gourmet program at midnight. You may feel as if the living room suddenly turned into a restaurant. It will entertain audiences, and benefit programs and commercial directors as well.
Is it possible to realize such a TV set? We interviewed Mr. Hattori, who is in charge of smell sensor development project.
We asked this question to Mr. Hattori. He thought for a while, and then said, "Currently it's a little difficult to make such a TV set."
However, he also considered the possibility of this solution.
Hattori：We are currently developing smell sensors, which detect smells and visualize (digitize) them. The further process depends on the TV manufacturers.
However, if we can visualize smells, I think it will be technically possible to generate the same smells by combining smell components based on the digitized data.
If a TV set or the remote controller could have a mechanism of combining smells and releasing them, a TV that generates smells might be feasible.
▲MEMS Semiconductor Type Prototype Smell Sensors under development by TAIYO YUDEN—Connected with a smartphone to digitize smells in a real time basis
Smell generation technology is a technology to blend several smell components and then to release the blended smell. Speaking of releasing smells, various products such as aroma diffusers and deodorants are already in the market. According to Mr. Hattori, those products can only release pre-fixed smells.
―― Deodorant products have been established as an independent product category. What is the difference between smell generation and deodorization?
Hattori：Technically speaking, deodorization is easier than smell generation. There are two methods for deodorization: one is to cover a smell with a stronger one. The other is to create an opposite smell to wipe out the existing one. The procedures are already established, and we now have patterns of how to deal with particular smells.
This is the reason why hundreds of deodorant products are already in the market. In contrast, generating smells means combining hundreds of thousand smells that exist in nature, which is more complicated and difficult.
▲Smell Sensor Development Site, TAIYO YUDEN R&D Center
Mr. Hattori and other development staff are working on smell analysis.
Smell information database in the cloud
Smell generation includes several processes: sensing, identification, and digitization. For digitization, pre-learning by AI plays an important role. He talked about his struggles in developing the process.
Hattori：There may be a misunderstanding that our smell sensor can detect any smells, identify them, and digitize them at once. In reality, pre-learning by AI is required to do it. Our smell sensor determines the types of smells by matching them with learned patterns.
▲Sensing and Identification (image)—Identification of an object by differences in the components
Hattori：For smell identification, the first step we carry out is to visit our customers to collect smell data and understand what smells they want to detect. In our lab, we recreate these smells and repeatedly teach those patterns to AI. When smell patterns are accumulated and a database is established, a new smell, which is detected for the first time, could also be identified as "one similar to the other" by matching to the database in the future.
However, this technology still has a long way to go. There are challenges for it across the industry.
Hattori：There is no data compatibility between sensors developed by different companies. The accumulated database can be used only within each company as of today. For example, each sensor reacts differently to the smell of an orange because of different methods in digitization. Standardization of smell data still be an important arguing point for establishment of future smell database.
―― That reminds us of the home VCR war between VHS and Beta.
Hattori：Currently, we have not reached the competition stage yet. No smell sensor products are in the market. We currently focus on suggesting solutions to our customers who have smell issues utilizing our smell sensor technology, rather than building specific smell database. Solving their issues together with the customers is an important first step for us toward commercialization.
In the future, when various competitors start to commercialize smell sensors, building smell database in the cloud, establishing a common smell database becomes meaningful. Then we will start talking about the standardization of smell data to maintain data compatibility.
▲Example of sensing an orange smell — "Smell of orange" is determined by the reactions of 16 channels in smell detection sensor
From 'Sensing' to 'Visualization' of Smell
―― Are there any promising fields or product markets with smell generation technology?
Hattori：Combination with deodorization technology could expand its applications. For example, an application to keep the room's smell in a same status could be possible, registering ideal air conditions in advance, constantly sensing the air status, and releasing an appropriate deodorizer or aromatics.
A robot cleaner with a smell sensor could spray deodorant in spots where it detects foul odors.
Demand in sharing services is also expected, which have been growing recently.
Hattori：There are various types of sharing services, such as car sharing and room sharing. Deodorization is an important factor in these businesses. Our smell visualization represents smells objectively and numerically rather than describing about the smell subjectively like "good smell" or "bad smell." That way, the area can be demonstrated as deodorized space by objective evidence.
▲In the development process, they also had to smell some bad odors!
Mr. Hattori explained the high level of interest in smell sensor with actual examples.
Hattori：Smell visualization has drawn attention from companies, which sell perfume products. They show interests in digitization of their perfume offerings developed by their perfumers.
The same interest consumers may have. By registering a favorite smell in a database, it may be possible to buy perfumes online. AI could recommend products based upon the customers favorites.
In the previous topic, I talked about monitoring human heath by sensing body odor. A pet robot with such a function could sense the smell of its owner and say something like, "You look a little tired today."
Smell generation products with vast potential
When smell generation technology is completed, commercialization of 'smell generating' products may become more widely used. Sharing smells with others could also become possible. Smell generation and smell analysis are paired technologies. Let's look forward to further improvement in smell sensor technology.
For more details,
please visit TAIYO YUDEN website.