The algorithm could be used for both games and films

Computer algorithm accurately detects human emotion

A computer algorithm has been developed that can read facial expressions and determine whether they denote happy, sad, angry or almost any other emotion with nearly 99 per cent accuracy.

The technology could be used by the games industry to capture and display player’s emotions on their in-game avatars or activate suitable actions to communicate with other players in various scenarios, including educational applications.

The developers, Hyung-Il Choi from Soongsil University, in Seoul, Korea, and Nhan Thi Cao and An Hoa Ton-That of Vietnam National University, in Ho Chi Minh City, said the system is simple and fast and has been tested on thousands of facial images.

Fundamentally, the system uses mathematical processing to measure eyebrow position, the openness of the eyes, mouth shape and other factors in order to correlate those with basic human emotions: anger, disgust, fear, joy, sadness, surprise and a neutral expression.

The system is even able to accurately determine the emotions of facial images that are just 48 pixels square.

Facial expression recognition has been the focus of much research in recent years, thanks to the emergence of intelligent communication systems, data-driven animation and intelligent game applications.

Last year, a humanoid robot named Han was revealed. It is equipped with pattern recognition software to get a reading on facial expressions and it is able to recognise and interact with the person in front of it.

"When facial expressions recognition of players is applied in an intelligent game system, the experience can become more interactive, vivid and attractive," the developers of the new algorithm said.

They believe that the system could be adapted in the future to track the emotional expressions of actors voicing the characters in animated movies and other media for more realistic real-time emotional expression.

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles