Odor Data AnalysisFeatured Technologies
February 15, 2018
Technology that brings to life the sense of smell―the most difficult hurdle in the reproduction of human senses
As IoT permeates society, more and more projects are underway to capture and visualize the movements and statuses of humans and objects as data. The accurate detection and understanding of the surrounding environment and circumstances, in addition to people and objects, is also gaining more attention to promote semantic comprehension of data.
In line with this trend, developments in alternative mechanisms to mimic human senses of vision, hearing, touch, taste and smell have shown progress—with only the reproduction of the sense of smell falling behind in terms of practical applications after more than 30 years.
Odor contains a lot of information, and its range of use is expected to be extensive. For example, it is known that human breath can reveal that person's diet, any deficiency in nutrients, and disease types.
Nevertheless, odor is made up of a mixture of odor molecules—there are said to be more than 400,000 different odor molecules—and as such, 'smelling' out the different odors with accurate detection is an extremely challenging task.
The Membrane-type Surface Stress Sensor (MSS), an olfactory IoT sensor developed on the initiative of the National Institute for Materials Science, now made possible the perception and identification of odor.
NEC's role in odor data analysis
As one of the seven founding organizations, NEC participates in the MSS Forum, which conducts experiments demonstrating the performance of olfactory IoT sensing systems―adopting membrane-type surface stress sensor―for their industrial standardization.
MSS is the world's first ultra-compact, high-sensitivity sensor that can be used in different applications to identify odors with a single device. This enables the conversion of odor into information (odor data) that can be interpreted by humans and machines.
NEC collects the odor data and measurement conditions data in a cloud and analyzes them using NEC's proprietary machine learning technology, Heterogeneous Mixture Learning. The Heterogeneous Mixture Learning automatically discovers massive patterns from big data and generates a prediction model for each condition. Additionally, higher-accuracy odor discrimination is realized by selecting the optimal prediction model according to the situation and considering the differences, such as measurement conditions and industry types.
Examples of odor data analysis application
NEC has demonstrated the use of odor data analysis in disease detection and health management based on breath and body odor.
Medicine and health care
NEC has demonstrated the use of odor data analysis in measuring the ripening level of fruit.
Food and primary industries
Applications in environmental testing and hazardous substance detection are anticipated.
Space and environment
Hazardous object detection, etc.
Safety and security
Product planning and development, warehousing management, etc.
Manufacturing and logistics
Supporting in determining fragrances in cosmetic products, perfumes, detergents, etc.
Fragrance and sence of smell determination