The software created by researchers at the Nanyang Technological University can tell the difference between people walking, talking and acting normally, and abnormal behavior such as a fight or someone collapsing.

The Singapore team recorded and classified 73 features of human movement, such as speed, direction, shape and pattern.

The features were then used with existing "neural network" software, which can learn and remember patterns, to create a new program.

"Each of the features is actually generated from a formula ... then the learning software will be able to classify certain motion as normal or abnormal," associate professor Maylor Leung said.

"It's something new. No one has tried (developing it) and so far we are successful," he said.

Images fed to the software, such as from a surveillance camera, are analyzed almost instantly and with 96 percent accuracy, Leung said.

The software can trigger an alarm when unusual movements are detected, making it well suited for surveillance.

Creating the artificial intelligence needed to recognize complex human motion has been a challenge, Leung said.

It is difficult for the human eye to accurately judge motion, such as speed, and even harder for a software program to do so, he said.

Leung is looking for partners to commercialize the software. The research, which took two and a half years, is pending publication in several technical journals.