National / Crime & Legal

Japan tapping gait recognition tech in criminal probes

JIJI

Gait recognition technology, a method to identify people by characteristics shown unconsciously in the ways they walk, is being utilized in criminal investigations in Japan.

The technology enables the identification of individuals even from images taken from a distance and low-resolution footage. According to advocates, a video image of only two strides is sufficient to identify a person with a high rate of accuracy, based on arm swings, length of stride and other characteristics.

Researchers are working to improve the accuracy of the technology with the use of artificial intelligence.

In a brazen daytime attack in Tokyo’s upscale Ginza district in April 2017, a man was robbed of some ¥40 million on a street after he had converted gold into cash.

The police collected footage from some 740 security cameras, according to informed sources. The gait recognition technology helped identify three of the four men arrested in the case even though they wore face masks or helmets when they committed the crime.

The technology can identify people even if they pretend to be others, because they “show their own unique gait patterns unconsciously,” said Hironori Yamauchi, a senior official of the Japan Image Analysis Association and a professor emeritus at Ritsumeikan University.

The National Police Agency’s National Research Institute of Police Science introduced the gait recognition system in 2013 on a trial basis and used it in investigations of some 60 cases. Analyzed data have sometimes been used as evidence during trials.

Yasushi Yagi, a professor at Osaka University’s Institute of Scientific and Industrial Research, which developed the gait recognition system, has come up with new technology that can identify with 96 percent accuracy a person walking in different directions in video images.

The new technology, announced last November, builds on an AI technology known as deep learning for processing large volumes of data. The AI learned gait patterns of some 10,000 people of various ages.

Yagi said he hopes to develop a system for tracking terror suspects based on gait patterns registered in advance using the new technologies, with the aim of employing it at large-scale events such as the 2020 Tokyo Olympics and Paralympics.

“In the future, I also want to apply the technologies to check health of people from the ways they walk,” he said.