Walk this way: software performs emotion recognition by your gait
A new article written by the managing partner of artificial intelligence analyst firm Cognilytica peers into yet another tricky nook of life alongside autonomous systems. What if a machine could read your mind by watching you walk?
Ron Schmelzer, Cognilytica’s managing partner and principal analyst, pushed the idea around in a contributed article to Forbes.com. The piece focuses on Aniket Bera, an assistant research professor at the University of Maryland’s Institute for Advanced Computer Studies.
Bera also is doing research with the Geometric Algorithms for Modeling, Motion and Animation, or “Gamma,” group. He focuses on social perception by intelligent agents and robotics.
In the article, Bera says robotics will continue to work more and more closely with humans, and to do that safely, algorithms have to do more than predict actions and reactions in co-botting environments — still a hugely demanding task.
Schmelzer writes that “robots need to understand human emotions, feelings, intent, social boundaries, and expectations.” Facial recognition does not go far enough in knowing (or at least intelligently guessing) what a person might do next.
Expressions are fleeting and deceiving, can reflect thoughts entirely unrelated to whatever task is at hand, and can be hidden by only a head turn. No need to tell this to anyone who has tried to read NBA star point guard Stephen Curry, who feints with his face first.
Better, Schmelzer says Bera is reading emotions based on analysis of a person’s gait. The Gamma group, now 25 years old, has created ProxEmo, an algorithm that reportedly can do just that.
A moving Clearpath Jackal robot running ProxEmo to interpret three-dimensional live video captured by an off-the-shelf camera has achieved a mean average emotional prediction precision of 82.47 percent on the Emotion-Gait benchmark data set. The group’s paper is here.
The list of potential uses mentioned in the Forbes article is long. An algorithm might foil a hijacking or spot a suicide in the offing.
And while Bera says privacy is protected by pixelating all identifying features but the gait, it could be argued that people, particularly those in developed economies, will dislike being “read” as they walk along with their thoughts as much as or more than facial recognition itself.
Article Topics
artificial intelligence | biometrics | emotion recognition | gait recognition | machine learning | research and development | robots
Comments