Fujitsu has developed a new behavioral analytics platform that can learn to identify specific human behaviors with minimal amounts of video data. Dubbed ‘Actlyzer’, the platform is able to recognize approximately 100 basic actions like head movements, hand movements, and walking patterns, which can in turn be used to train the AI to recognize more intricate behaviors.
To that end, Actlyzer essentially uses basic actions as building blocks, with Fujitsu noting that a complex action represents a series of much simpler activities. By selecting the right sequence of actions, Actlyzer customers can teach the platform to look for the complex actions that are most relevant to their own business. Fujitsu boasts that the training process can be carried out in less than a day, giving customers much faster turnaround time when using the technology.
Security is the most obvious application of the technology. Organizations can instruct Actlyzer to watch for suspicious individuals, such as people scoping out locked doors in restricted areas.
However, Actlyzer can also be deployed to gather information and improve an organization’s analytical capabilities. For instance, retailers can use it to figure out which items customers are reaching for on the shelves, while employers can use it to make sure that critical work procedures are being followed.
Fujitsu will launch Actlyzer in Japan before the end of the 2019 fiscal year, and will then expand to international markets through the Fujitsu Human Centric AI Zinrai portfolio. The solution seems to be geared towards analytics rather than personal identification and authentication.
Fujitsu still seems to prefer palm vein technology in that regard. The company recently upgraded its AuthConductor authentication platform to extend the utility of the solution to workstation and PC logins. It has also partnered with ImageWare to protect its RunMyProcess Digital Suite, which can be used to develop applications for IoT devices.
November 25, 2019 – by Eric Weiss