Global Site
Breadcrumb navigation
Recognition AI
AI TechnologiesJune 30, 2022

Recognizing Every Type of Object and Person with the Goal of Building a Digital Twin
We are in a era in which COVID-19, environmental problems, and the global environment are changing at a rapid pace. It is becoming increasingly important to quickly predict and respond to changes. An innovation which is attracting attention for helping to predict such changes is the "digital twin." NEC's Recognition AI research is currently working to realize the digital twin through the development of technologies to digitize the world. Digitizing and connecting every object to recreate cities and transition to a world that can be simulated. In addition, we are aiming to become able to observe changes in people and society by also digitizing "people."
As represented by the image recognition work started in the 1960s and the face recognition shown to have the highest accuracy in the world by NIST*, NEC is a company with significant achievements in recognition technology. Utilizing know-how accumulated over many years, we are advancing cutting-edge research that is attracting worldwide attention.
*NIST: National Institute of Standards and Technology
Robust and lightweight recognition AI predicated on operation in a real environment
NEC's recognition AI is uniquely effective in a real environment. Naturally, it displays robust recognition performance that is resistant to noise and environmental changes as well as achieves realistic processing speeds. Assuming that every object becomes a digitized digital twin, it will be impossible to run a high performance GPU server to recognize individual objects. We are developing technologies that can deploy lightweight and realistic solutions while utilizing edge AI, etc.

Primary research technologies
Biometric authentication (face recognition/iris recognition/voice recognition)
NEC's biometric authentication features not only accuracy and speed but is also resistant to changes over the years and environmental changes. Recently, we have also been working on high accuracy multimodal biometrics which combines two or more recognition technologies. We are researching not only mathematical approaches but also basing our biometric authentication on a wide range of knowledge including neuroscience.
Cancer detection with an endoscope
This technology identifies cancer from an endoscope image. It not only detects cancer but also the degree of cancer progression.
Insole type sensors
These are sensors which integrated into the insoles of shoes. They comprehensively analyze one's gait in terms of step size and angle of swing, etc. We are pursuing practical application in hospitals and rehabilitation sites.
Using NEC’s Gait Analysis Technology to Promote a Healthy Walking Posture: A-RROWG Walking Sensing Insoles

Optical fiber sensing
It utilizes optical fiber laid for communication as sensors. This technology is able to read the vibrations, temperature, and sounds of each point and analyze the traffic volume and infrastructure degradation, etc. We are advancing research aimed at application in smart cities and autonomous driving.
SAR (Synthetic-Aperture Radar) based surface observation
This technology illuminates the ground surface with microwaves from SAR satellites such as "ASNARO-2" and aircraft to observe the surface by analyzing the reflected waves. Because it is not affected by clouds or nighttime conditions, it is able to obtain information in a stable manner. We are advancing the use of this technology for the observation of ground settlement and infrastructure deterioration as well as agriculture.
Two-dimensional Small Displacement Analysis Technology Utilizing Satellite Radars Enabling the Deterioration Inspection of Multiple Roads and Buildings in Urban Areas

Instant object registration
This technology can register new objects in an image recognition model just by rotating the objects in front of a camera. It automates work such as learning data cleansing and annotation that was previously required when registering data in an image recognition model. We are promoting the utilization of image recognition in stores and warehouses, etc.
Significantly Reducing the Time and Effort to Register New Products
Instant object registration for image recognition

Liquid product inspection
This technology automates the external inspection of liquid products (cancer medications, vaccines, etc.) which previously relied on the human eye. By capturing the drifting motion with a high-speed camera, AI can recognize even small objects of about 50 microns that were difficult to identify as bubbles in a liquid with conventional image inspection systems.