New webcam-based AI system reads and responds to human facial expressions

A doctoral project at Tallinn University has developed the Enactive Facial Expression Pipeline (EFEP), which uses a standard webcam to analyze human facial expressions and animate a digital character's response in real time. Unlike systems that assign broad emotion labels to static images, EFEP treats facial behavior as an ongoing interaction, using individual facial muscle movements (action units) for finer control. The research combined existing datasets with a new collection of 60 participants reacting to videos in five imagined social contexts. A key finding was that three different facial-expression analysis tools (iMotions, Mini-Xception, and Py-Feat) disagreed on emotion predictions for the same recordings, highlighting methodological problems in emotion recognition. The project also addressed demographic bias, showing that models trained on limited groups performed unevenly, and that diverse training data reduced but did not eliminate bias. The system is intended for interactive media and creative applications, not medical or employment assessments, and the researchers emphasize that a responsive digital face does not equate to true emotional understanding.

(Source：The Brighter Side of News)