Fujitsu Unveils Sophisticated Deep Learning Model for Behavior Recognition

Fujitsu has announced a new AI system for behavior recognition that may offer additional biometric applications.

The technology is based on a deep learning model that has been trained on unusually complex skeletal models of the human body.

As Fujitsu explains, in understanding the movements of a human body, its AI system has been trained on the movements of more joints than are used in training conventional behavior recognition systems. For example, whereas a standard AI system will look at the movement of an elbow in a certain gesture, Fujitsu’s system looks at how the elbow, hand, and shoulder are moving in tandem, which allows it to get a better understanding of the movement being undertaken.

Potential applications of the technology highlighted by Fujitsu include assessing manufacturing procedures, or automatically detecting unsafe behavior in public spaces, among others.

This technology could also offer potential applications in gait recognition and related biometrics. Still a somewhat nascent area in the field of biometrics, gait recognition has been seriously explored by the likes of Mastercard and the Department of Defense. Fujitsu’s technology could allow for more enhanced gait recognition technology, and may be used to extend the concept to other behavioral movements that might be used for the purposes of authentication or identification.

Fujitsu made no mention of such biometric applications in announcing its technology, but it’s worth noting that the company is actively involved in the biometrics space.

In any case, going forward, Fujitsu indicated it will integrated the AI model into its Actlyzer behavioral analysis platform.

The company announced its sophisticated deep learning technology during this week’s International Conference on Pattern Recognition, which will conclude on January 15.

–

January 13, 2020 – by Alex Perala