Global Site
Breadcrumb navigation
Message for students 2025:Masashi Usami
March 19, 2025
Estimating health conditions and inner states from facial videos

Masashi Usami
Researcher
Biometrics Research Laboratories
Usami joined NEC in 2021 after his doctoral program in experimental particle physics. He engaged in research on fairness in facial image analysis and machine learning, and his research paper was accepted at the 2023 IEEE International Conference on Image Processing (ICIP 2023), a top-tier international conference in the area of image analysis. He participated in launching a research team dedicated to more actively leveraging facial information in healthcare and other fields in 2023. He is working on research that estimates health conditions and inner states of people using smartphones, with the aim of extending the healthy lifespan of society as a whole.

Attracted to an Environment For Proactive research
I majored in experimental particle physics in graduate school, so I was engaged in work that was quite close to basic research. When I started thinking about my future, I began to wonder if there might be a way to use what I've learned to be more helpful to society. While the knowledge from my basic research might not be directly applicable right away, I believe there are ways I can still contribute with what I've learned. I still believe that pursuing truth guided by my own intellectual curiosity is a great thing, but dedicating my whole life to it just didn't quite feel right for me. That's when I started looking into corporate research as another option.
What solidified my decision to apply to NEC were the laboratory tour and other exchange events. During these events where students and NEC researchers presented and discussed their research topics, I noticed that the researchers I spoke with were very enthusiastic and open. I liked the atmosphere they created. They talked about their work with genuine enthusiasm, and it was clear they were truly passionate about their research. I was really impressed by their proactive and joyful approach to conducting research.
During the interview, they surprised me by saying something like, "You’re welcome to join us, only if you want to." That really resonated with me. They said, "We’d love to work with you, so if this is where you want to be, come join us. Of course, we'll support your decision, even if you choose another company. But if you decide to come here, let's work together." Some other companies where I interviewed expressed a strong desire for me to join them, for which I was very grateful. However, none had the same approach as the researchers at NEC. Each researcher there is independent, respects others, and pursues their research interests passionately. The attitude I experienced at NEC could only come from such individuals, and that is why I chose NEC.

Tackling Challenge of Fairness in Facial Image Analysis
When I first joined NEC, I was assigned to the facial image analysis team. During my interview, I expressed my desire to be involved in biometric authentication, one of NEC's strengths, which may have been taken into consideration. You might wonder how I managed the sudden transition from particle physics to image analysis, but I personally did not experience a significant gap. Whether it's AI or physics, the fundamental knowledge comes from mathematics and other general education skills. Naturally, I needed to learn new things, but that's expected when tackling something new. Changing your area of study can feel similar to transferring to another lab. Additionally, as a new employee, you receive support from your supervisor and mentor. I was given a sufficient period for thorough training when I first joined NEC.
My initial research assignment was to ensure fairness in facial image analysis. For example, when estimating a person’s attributes (such as race, gender, or facial expression) from facial images, unexpected biases may arise depending on the AI model used, such as decreased prediction accuracy for one gender. Such biases are challenges that must be addressed because they may potentially lead to discrimination if improperly implemented in society. It is a critical issue that NEC, involved in biometrics, must confront directly. However, ensuring fairness is challenging, as there are no universal or perfect metrics for many real-world problems that guarantee absolute fairness. I tackled this foundational research theme alongside my mentor.
What I focused on was the statistical bias in the training data, which is the source of the kind of bias mentioned earlier. Using the previously mentioned example, potential causes for differences in prediction performance based on gender include pre-existing gender imbalances in the vast training data or a large number of individuals with specific facial attributes for each gender. To address this, I developed an algorithm that robustly handles these data imbalances and can perform a fair estimation of facial attributes. When I shared this idea with my mentor, he suggested that I write a research paper because he found it interesting. I organized the study into a paper, which was submitted to and subsequently accepted by the 2023 IEEE International Conference on Image Processing (ICIP 2023), a top-level international conference in the area of image processing.
When a corporate researcher writes a research paper, it can sometimes be perceived as part of business promotion. However, I personally view writing a research paper as an important opportunity to deepen my understanding in that field. Completing this research paper gave me, for the first time, the confidence to call myself a researcher in the study of facial attributes.

Using facial information for daily healthcare
From my third year at NEC, I began researching how to understand health conditions and inner states by actively using facial information. This direction was inspired by my supervisor's advice; he suggested that since my work involves using facial information and attributes rather than face recognition itself, pursuing applications in areas like healthcare could be more fitting. I agreed with his insight and joined a new team led by Ms. Umematsu from our laboratory.
Our team engages in both individualistic and corporate research, collaborating with business unit members to present demonstrations at exhibitions. This is also interesting in its own way.
We are currently developing a preliminary demonstration for research purposes that uses smartphone-captured facial images to estimate indicators such as pulse rate, respiratory rate, SpO2, and other metrics associated with inner states. By integrating insights from future collaborative research with universities and university hospitals, I hope to develop an application that allows individuals to detect changes in their health conditions on their own. Our primary mission is to extend the healthy lifespan of society as a whole through the ubiquitous use of smartphones.
NEC Laboratories attract researchers with a variety of specialties and backgrounds. In addition to biometric authentication and healthcare, which are my areas of focus, NEC engages in diverse research topics such as natural language models, telecommunications technologies, and quantum computing, comparable to those found at universities. I am currently enjoying my research, but even if I had encountered themes that didn't suit me, I believe I would have found at least one area that fit my interests and allowed me to work proactively.
At NEC, being directed by a supervisor to pursue a specific research topic in a specified manner is quite rare. It is important to independently find and pursue topics of your own choosing and interest. Naturally, such freedom is accompanied by responsibility. I believe NEC offers an excellent environment for those who are fully aware of and can manage these responsibilities.
This is why I encourage students to consider their career paths based on their desire to conduct specific research, rather than being constrained by the categorization of corporate versus academic research. Consider whether you can undertake that research and find joy in it—this will lead to a fulfilling research life.
- *The information posted on this website is the information at the time of publication.
A day at work


Message to my past self in my school days

Private column
Since I started healthcare research, I've been paying more attention to my health condition and have started preparing my own meals with consideration for nutritional balance. I've collected various cooking utensils and have tried making dishes like stew and stuffed cabbage. On weekends, I treat myself by buying and cooking delicious meat. Since the beginning of this year, I've also been reassessing my exercise routine. For example, I installed NEC’s health management app, WoLN*, to measure my steps, and I joined a gym and got guidance from a trainer. I'm enjoying my workouts!
