Asilla, a developer of security systems with world-class behavior recognition AI at its core, launched the Human Science AI Research Team (hereinafter referred to as HSAR) in February 2023, a specialized team centered on human science research.
In this article, we interviewed Mr. Masahiro Wakasa, CTO of Asilla, the founder of the team, about how "gait analysis," one of Asilla's core technologies, and HSAR's research are connected.
Gait analysis is a technology that focuses on the so-called "human gait. It is a field that requires in-depth knowledge of the human gait, as it is the research and development of technology that identifies "who the person is" based on characteristic movements.
At present, Asilla is conducting two main research projects in gait analysis: "gait recognition" and "trajectory analysis.
The "gait recognition" is a technology that identifies people using the individuality of the way they walk. It is the only technology that can authenticate individuals by recognizing the way they walk, even at locations where their faces cannot be identified in the video image.
For example, with security cameras, depending on the distance and angle of view, it may be difficult to identify a person because their face cannot be recognized, but the Walking Pattern Recognition System can identify a person because it can recognize the way they walk even from 50 to 100 meters away. Another feature of the system is that it can identify and track a person even between different cameras because it focuses on the way he or she walks.
As a biometric authentication technology that uses human physical characteristics to authenticate individuals, "gait recognition" is also attracting attention for its potential to become as accurate as facial recognition in the future. It is also expected to be used in crime prevention efforts and layout design in urban development.
アジラで行っている「軌跡分析」とは、人が歩いた道筋（軌跡）から、その人の心理状態の特定や今後起こりうる行動を推��する研究です。
例えば、通常の人の場合はまっすぐ歩く傾向にありますが、ティッシュ配りやキャッチセールスなど一般の動きとは異なる人物の場合、何回も同じエリアを行き来したり、人に話しかけるというように、動きに違いが生まれますよね。軌跡分析では、このような動きの違いをAIが自動的に検知・抽出するというアプローチとなります。
また、「歩容分析」は「歩行」に着目しているため時間軸が数秒と短いのに対し、「軌跡分析」は「道筋」に着目しているため10秒前後と長期的な動きを対象としていることが違いのひとつです。AIの業界では、正常行動と異常行動を区別し検出する「Anomaly detection（異常検出）」というアルゴリズムが存在しており、弊社でも現在研究中で、プロダクトへの展開も検討しています。
このようにアジラでは、「歩容認証」と「軌跡分析」の２つのアプローチから「歩行解析」の研究を深めており、更にAIの検知・推測の精度を高めるべく研究開発を推進しています。
As reported in the article on the launch of HSAR, in gait analysis, Asilla has been conducting research from two axes: traditional "computer science" as well as "human technology" to deepen our understanding of people themselves.
Until now, our research has focused on the "typical computer science approach," in which AI is given learning data on various human walking styles, and AI learns automatically. However, there was a concern that this approach might eventually reach a plateau, even though it could improve accuracy to some extent, because of the risk of the AI automatically learning with incorrect parameters.
Therefore, HSAR incorporates the knowledge of human science into the process of "assigning the points (parameters) that AI should originally look at" from this perspective. It is like the findings of human science becoming "the right guideposts for AI.
For example, for a male in his early 30s, his walking speed is about 5.7 km/hour, and the average stride length is about 70 cm. By adapting such "parameters to be viewed from human science" to AI, Asilla aims to further improve the accuracy of AI.
まだ感覚値ではありますが、成果はすでに見え始めていると感じています。そ���中でも「これまでAIの内部でブラックボックスだった事が少しずつ可視化できつつある」ということがひとつ大きな成果と感じています。
どういうことかと言うと、これまでのAIは予測精度を高めるために複雑に構成されているため、判断基準や根拠が不明確である「ブラックボックス化」が問題となっていました。ただし、AIにヒューマンサイエンスの視点が新たに加わったことで「このAIは歩き方の”この部分”に着目し判断した」というAIの判断軸や根拠が分かるようになりつつあります。
例えば、本来は肘の角度を見て判断すべきにも関わらず、AIは頭を見て判断してしまっていることがありました。その後、AIに間違っている点を指摘し「ヒューマンサイエンス的に見るべき視点（パラメーター）」を付与したことで、結果が改善し精度向上したケースもありましたね。
こうした「AIの見える化」により、ヒューマンサイエンスとAI双方の視点（パラメーター）が一致する整合性の取れた有力なモデルが発見できたり、反対にAIが見るべき視点と全く異なるパラメーターをみて判断している再評価対象のモデルを見つけることが出来たりと、きちんと選定できるようになりつつあります。
While the introduction of human science is expected to further improve the accuracy of AI, we believe that there is a possibility that AI beyond the understanding of human science will emerge.
For example, even though the head is not considered an important parameter in current human science, AI may lead to new discoveries that have not been considered in existing research, such as the fact that the way the head moves has a significant impact on age and gender.
In fact, the phenomenon of "AI surpassing humans" has already occurred in other industries, with a familiar example being recommendation systems. The same may be true in the future for human behavior, which is the subject of our work. It is exactly as if AI is going to surpass humans.
While keeping such a possibility in mind, Asilla aims to "develop AI with a good balance between humans and AI" by utilizing human science as "knowledge to improve the accuracy of AI," rather than thinking that "human science is 100% correct. We aim to "develop AI that balances both humans and AI.
Based on the philosophy of "Technology Driven Future," Asilla conducts research and development on a daily basis in order to realize a world in which all people can live safely and securely. In the field of behavior recognition AI, in particular, we have a strong desire to become a world leader, and our researchers and members are working together as a team to constantly evolve.
In the future, we will further promote the further evolution of our behavior recognition AI technology and strive to create an environment where all our engineers can pursue their own careers and always be excited about their work. If you are interested in AI related to human behavior, we look forward to working with you to create something together.
Inquiries about HSAR: pr@asilla.jp Contact: Wakasa
After completing his postgraduate studies at Tokyo Institute of Technology, he worked for JGC Corporation where he was involved in plant design IT work on overseas construction projects. He then joined Asilla Corporation, where he was in charge of proof-of-concept and product development projects related to behavior recognition AI, and became the company's executive officer and CTO in 2022. He continues to focus on product development using AI technology and research and development of new technologies.
