Breadcrumb navigation

Smells Revolutionary: Explaining the sense and sensibility of scents

We are surrounded everyday by all sorts of scents and odors, which carry a surprising amount of information. This treasure trove of information, however, has been kept locked, preventing us from understanding its inner secrets.

This treasure trove of information, however, is slowly being unlocked. And the key to unlocking it is the advanced ”MSS Olfactory IoT System,” which was developed through breakthrough olfactory sensors and advanced AI technologies. Let’s take a look at initiatives aimed at developing and commercializing new mechanisms for analyzing olfactory data to easily and accurately distinguish scents and odors.

Hints in analyzing odors from the functions of the nose and brain

Despite being replete with a vast amount of information, odor has remained inaccessible. Aiming to unlock the information within odor for application in various areas, many companies and research institutions have been attempting to develop olfactory sensors for over 30 years; but no one has so far commercialized an olfactory sensor for everyday use. The reason lies in the complexity of scents and odors.

There are 400,000 types of odor components found in nature. And one particular scent is made up of a few hundreds to a few thousands of odor components. For example, there are 500 different odor components responsible for the smell of coffee and 1000 odor components found in human breath.

NEC Data Science Research Laboratories Senior Specialist Junko Watanabe

”Among the five senses, the physical signals from sight and hearing are easier to measure than the chemical signals from taste and smell. In particular, due to the many types of odors, there had been no standard for measurement. Also, since it is easily affected by environmental changes such as temperature and humidity, it had been difficult to measure odor.” (Watanabe)

The fusion of sensing technologies that detect different types of odor components with high sensitivity and AI technologies that are capable of highly accurate analysis has significantly increased the prospect of commercializing systems for analysis and identification of scents and odors.

First let’s look at the mechanism of how organisms sense odors. For example, humans detect odor components through the nose and transmit them as electrical signals to the brain. The brain then analyzes the combinations of signal patterns to distinguish differences in odors.

The ”MSS Olfactory IoT System” was developed using this biological mechanism as a hint. This system is composed of the ultra-compact Membrane-type Surface stress Sensor (MSS) developed by the National Institute for Materials Science (NIMS) and the odor data analysis system developed by NEC using AI. Simply put, the MSS functions like the nose, while the odor data analysis system functions like the brain.

In September 2015, the MSS Alliance was launched to accelerate the practical application of the MSS Olfactory IoT System. Several companies possessing specialized technologies and knowhow in different fields joined the alliance to help in the mass production of chips and provide the reference gas for measurement, etc. The alliance has pursued the construction of platforms for the establishment and standardization of basic component technologies and the development of measurement modules embedded with the MSS chip, among other initiatives.

zoomLarger view
Method for analyzing sensor data based on the heterogeneous mixture learning technology

NEC’s AI technology enables automatic and highly accurate analysis of odor

Heterogeneous mixture learning technology, one of NEC’s leading AI technologies, is at the core of the development of the odor data analysis system, which functions as the brain in processing scents and odors. Conventionally, testing was done by manually setting each condition for discriminating odor, wherein the process not only costs time and labor, but also lacks discriminatory power.

Heterogeneous mixture learning technology uses the wave profile of all electronic signals received to accurately and automatically create a model for discriminating odor. This allows dramatically improving the accuracy and speed of discrimination. Moreover, an excellent advantage of heterogeneous mixture learning technology is its ability to discriminate odor based on the characteristic features of wave patterns of electronic signals, even when odor components and mechanisms are unknown.

Ms. Watanabe, who plays a dual role of research and business creation management and works with researchers, in-house business divisions, as well as partner companies in promoting the research project on the odor data analysis system, shares the following inside story about the development of the analysis system.

”First, in order to apply the heterogeneous mixture learning technology we had to gather various types of odor data for testing and conduct different experiments.

For example, we tried some ”life-risking experiments” by measuring odor data from urine samples obtained from our young researches who engaged in heavy drinking the previous day and from those who only had water (laughs). Establishing a theory requires overcoming many challenges, but it is well worth all the hardships.”

Potential applications in various fields, including medicine, healthcare, and environment

Compact and portable, the MSS Olfactory IoT System can be easily used anywhere. Once commercialized, it has potential for various applications in various fields. For example, it can be used in medicine and healthcare for checking dietary habits and state of health from exhaled air or body odor.

It can also be used in environmental monitoring to discriminate harmful substances and detect offensive odors in living spaces, in security surveillance to detect explosive materials and deterioration of facilities, in fragrance research for cosmetic products and detergents, in determining ripeness of fruits, in livestock health management, and a wide range of other possible applications.

Embedding the MSS chip in wearable devices and smartphones will make the management of daily eating habits and state of health from the odor of sweat and breath, for example, an everyday thing. That day may soon come!

Palm-sized measurement module embedded with the olfactory sensor (MSS).

(February 28, 2020)

  • facebook
  • Linkedin