Gait recognition: AI can recognize you based on how you walk

• Author:
• Author name

  Quantumrun Foresight
• February 1, 2023

Even the way people walk can be used to identify them, like a fingerprint. An individual's gait presents a unique signature that machine learning algorithms can analyze to recognize a person from an image or video, even if their face is not in view.

Gait recognition context

The most common type of gait study is the processing of temporal patterns and kinematics (the study of motion). An example is knee kinematics based on different marker sets on the tibia (a leg bone), computed by segmental optimization (SO) and multi-body optimization (MBO) algorithms. Sensors such as radio frequency (RFS) are also used, which measure bending or flexing. In particular, RFS can be placed in shoes, and communication data sent to a computer via Wi-Fi to detect dance movements. These sensors could track the upper and lower limbs, the head, and the torso.

Modern mobile phones are equipped with various sensors, such as accelerometers, magnetometers, inclinometers, and thermometers. These features allow the phone to monitor the elderly or disabled. Additionally, mobile phones can identify hand movements during writing and subject recognition using gait movement. Several apps also help monitor physical movements. 

An example is Physics Toolbox, an open-source app on Android. This program allows users to access various sensors, which include a linear accelerometer, magnetometer, inclinometer, gyroscope, GPS, and tone generator. Data collected can be displayed and saved as a CSV file on the phone before being sent to Google Drive (or any cloud service). The app's functions can select more than one sensor to gather various data points simultaneously, resulting in highly accurate tracking.

Disruptive impact

Gait recognition technology creates an identification by matching a person's silhouette, height, speed, and walking characteristics to information in a database. In 2019, the US Pentagon funded the development of smartphone technology to identify users based on their walk. This technology was distributed widely by smartphone manufacturers, using sensors already in the phones. This feature ensures that only the intended user or owner can handle the phone.

According to a 2022 study in Computers & Security journal, every person's way of walking is unique and can be used for user identification purposes. The objective of gait recognition is to authenticate users without an explicit action, as related data is continuously recorded while the person walks. Therefore, transparent and continuous smartphone protection can be provided using gait-based authentication, especially when used with other biometric identifiers.

Aside from identification, healthcare providers can use gait recognition to monitor their patients remotely. A posture analysis system can help diagnose and prevent various deficiencies, such as kyphosis, scoliosis, and hyperlordosis. This system can be used at home or outside medical clinics. 

As with all recognition systems, there are concerns about data privacy, particularly biometric information. Some critics point out that smartphones already collect too much data from users in the first place. Adding even more biometric data can result in people completely losing their anonymity and governments using the information for public surveillance.

Implications of gait recognition

Wider implications of gait recognition may include: 

• Healthcare providers using wearables to track patient movements, which can be helpful for physical therapies and rehabilitation programs.
• Sensors being used for elderly-assistive devices that can monitor movements, including alerting nearby hospitals for accidents.
• Gait recognition being used as an additional biometric identification system in offices and agencies.
• Smart devices and wearables that automatically delete personal information when they sense that their owners are no longer wearing them over a certain period.
• Incidents of people being wrongfully arrested or interrogated using gait recognition evidence.

Questions to comment on

• How else do you think companies will use gait recognition technologies?
• What are the potential challenges of using gait as an identifier?