AI system detects ‘out of place’ behavior

J-STORIES - The Kabukicho area of Shinjuku, Tokyo is packed day and night with visitors who come for a variety of purposes, such as shopping, entertainment, or business, but the area has also long been known for its ties to the darker sides of Japanese nightlife. One company has developed a system that it believes will make the district a safer place.

The Tokyu Kabukicho Tower, a skyscraper that opened this April, is equipped with a cutting-edge security system that uses AI to detect suspicious behavior and dangerous situations that might lead to crimes or accidents.

Video demonstrates how the Asilla AI system detects staggers and falls. Source: Asilla

The system has been provided by Asilla, a Tokyo company that develops behavior recognition AI. Its AI security system, also called Asilla, autonomously learns the behavior of people caught on security cameras, then immediately notifies the user of any “out of place” behavior that might precede criminal acts or accidents, thus preventing those incidents before they occur.

Video shows how Asilla detects "abnormal" behavior Source: Asilla

A wide range of subjects can be analyzed, such as falls, fights, vandalism, people unsteady on their feet, or who stay in one place for a long time. As well as unusual behavior, it can also look for lost children or obstructions that might affect accessibility for wheelchair users. And because it can look at flows of people within a facility, it can be used to analyze customer shopping habits or detect habitual shoplifting.

According to the company, the Asilla system is among the best in the world in terms of display speed, tracking ability, and accuracy when estimating a person’s posture from images and video.

The AI system analyzes images, detects suspicious or abnormal behavior, then immediately notifies the user.     Source: Asilla — The AI system analyzes images, detects suspicious or abnormal behavior, then immediately notifies the user. Source: Asilla

“To date, security cameras have often been used for verification after an incident or accident has taken place,” says Suzuki Yuka, a PR Manager for the company. “But Asilla detects behavior before it happens by noticing things that are out of place, then prevents it happening by reporting it to various devices within one second.”

According to Suzuki, Asilla has a number of advantages compared to products from domestic and overseas competitors. For example, it is able to detect various behavior patterns that are specific to humans, so there are almost no false alarms that might be set off by animals or inanimate objects. Furthermore, AI’s autonomous learning improves Asilla's accuracy without the need for human intervention and one server can process images from up to 50 cameras, which can lead to a significant reduction in surveillance costs.

The system is currently in use not only in Kabukicho, but also at universities and hospitals, and large commercial facilities such as the Shin-Marunouchi Building in front of Tokyo Station.

Asilla has been used to reduce risk to members of the public, such as by quickly detecting children climbing over handrails or running the wrong way on escalators.     Source: Asilla — Asilla has been used to reduce risk to members of the public, such as by quickly detecting children climbing over handrails or running the wrong way on escalators. Source: Asilla

The crime rate in Japan is low compared with other nations around the globe, but recent attacks on two Japanese prime ministers have raised questions about the country's current safety and security policy. In July 2022, then Prime Minister Shinzo Abe was shot dead while speaking at a political event in Nara, while last month current leader Fumio Kishida was the target of a potentially life-threatening incident at a fishing port in Wakayama Prefecture.

And with Japan set to hold the G7 meeting in Hiroshima this month, Japanese society has become more sensitive to bolstering public security, especially for visiting and other dignitaries.

And despite recent concerns about the dangers of AI technology, including an open letter penned by AI experts and industry executives urging the suspension of AI development due to the potential risk to society, it could hold the key to future security and safety endeavors.

One major worry relates to privacy, and to address such concerns Asilla has positioned its system as “a technology for realizing safe and secure spaces.” It has also established its own AI Charter that stipulates compliance with local laws in order to ensure the protection of privacy and maintenance of ethical standards. The company has also set up an expert committee to consider such matters.

This March the company started a joint project with Nagoya University to analyze people’s emotions from their gait. They are now working to develop a product that detects and prevents crime by analyzing emotional state, such as being agitated and stressed, and the company hopes to implement the technology in a security system this year.

Suzuki told J-Stories, “Going forward, by analyzing and researching the behavioral patterns of people around the world with different lifestyle habits and cultures, it should be possible to customize the system for each location.”