Hitoshi IMAOKA
Yusuke MORISHITA
Akihiro HAYASAKA
Facial recognition technology has attracted a lot of attention with the growing concern about the safety and security since the 9.11 terrorist attacks. Facial recognition is the most natural form of biometric recognition method for humans to practice. However, it is very difficult to achieve high accuracy against illumination variation or aging changes. Addressing this problem, the awardees started to study the core technologies of facial recognition such as face detection, facial feature point detection, and face matching (Fig.1) from 2002 and continuously collected large scale facial data. In particular, the awardees developed a metric learning method for the extreme multi-class classification problem(1). This break-through remarkably reduced the recognition error rate.
The awardees participated in the facial recognition evaluation program in 2009 hosted by the National Institute of Standards and Technology (NIST) to evaluate the effectiveness of their technology. NEC achieved the top ranking with the overwhelmingly lowest recognition error rate in a benchmark in 2010. This error rate was less than 1/10 compared with the No. 2 participant. After this achievement, NECfs facial recognition products were deployed world-wide. In addition, the latest evaluation report was published by NIST in September 2019(2). According to the report NEC achieved the top ranking for recognition accuracy and search duration, and as a result NECfs technology ranked No.1 in NIST benchmarks for the fifth time. 49 organizations including many top AI vendors in China, Russia and the US participated in this latest evaluation. NEC achieved a recognition error rate of 0.4% for 1.6 million enrolled persons, which is 1/3 compared with the No. 2 vendor. Furthermore, NEC also achieved a search duration of 230 million matchings per second (Fig.2). NECfs technology keeps the No. 1 in the NIST benchmarks with recognition accuracy in the decade since 2009 and received acclaim for practical utility for its robustness against aging and multi-ethnic groups.
NeoFace®, which is the facial recognition product developed by concentrating this technology and knowhow, has been recognized for its effectiveness because it can reduce the burden and error checking with eyes and can perform highly accurate authentication in a short time. This product is used in 45 countries or more for immigration systems and mobile device protection, and so on. It has dramatically changed authentication processes. This product was also adopted as the facial recognition system for immigration control at JFK International Airport in the US (Fig.3). It is essential technology for operating an airport safely and smoothly and for improving the customer experience. In the national identification system in India, this product is used by 1.3 billion persons as multimodal system including facial recognition. It serves as social infrastructure in India. In addition, this product has been adopted by the Star Alliance which is the worldfs largest global airline alliance, and has also been adopted as the main security systems of governments in England, Australia and Brazil. In this way, this product is used as the infrastructure of public safety systems around the world. In Japan, this product has been adopted as an anti-scalping system at concerts of 100 thousand people, and has also been adopted as the One ID system at Narita Airport where persons pass through from check-in to boarding by facial recognition. In addition, it will be used at security gates for staff for the Olympic and Paralympic Games Tokyo 2020 (Fig.4) .
Facial recognition will be adopted for stricter personal authentication like payments because of the significant improvement of recognition accuracy. It will become more and more common as a tool for achieving convenience and equity.
Because of such achievements, one of the awardees gave a keynote speech at the IEEE International Joint Conference on Biometrics 2014, which is a top conference in biometrics. The awardees also received many prizes such as the gThe Minister of Education, Culture, Sports, Science & Technology Awardh at the 3rd JATES Technology Management & Innovation Awards. In addition, the product developed by the awardees received the Frost & Sullivan Best Practices Award 2016, the CEATEC AWARD 2017 the Minister of Economy, Trade and Industry Award, and the Best 10 New Products Awards 2018 sponsored by Nikkan Kogyo Shimbun. Thus, the awardees well deserve the IEICE Achievement Award by virtue of these outstanding achievements.
1NEC is the Gold Partner for the Tokyo 2020 Olympic and Paralympic Games in the categories of "Specialist Public Safety Equipment & Software" and "Network Equipment".
