Face-recognition technology used in new lie detector test

Bradford University researchers have developed a system using cameras and computer software to spot small facial clues that someone is being dishonest.

The system picks up indicators such as lip-biting, slips of the tongue, nose-wrinkling and blinking. A thermal imaging camera is also used to measure flushing and blood-flow patterns around the eyes.

In future the system could be used covertly by police and border officials questioning suspected criminals or terrorists.

With hidden cameras positioned up to three metres away, suspects would not know they were being subjected to a lie-detection test.

Scientists at Bradford University hope soon to trial the technology at a UK airport.

Conventional “polygraph” lie detector tests involve wiring up subjects to equipment that measures blood pressure, pulse, heart rate, respiration and electrical skin activity.

However, people have been known to “beat” the test using a range of techniques, such as controlling breathing and heart rate. Professor Hassan Ugail, who heads the Bradford University team, said: “Our aim was to develop a purely non-invasive lie detector technology.

“We assumed this could be used in a covert situation, where the person we are monitoring potentially knows nothing about it. We don’t believe the polygraph is a 21st century technology.”

He described how the system worked yesterday at the British Festival of Science taking place at Bradford University this week.

“When a person is being deceitful, when a person lies, there is increased brain activity and this is reflected in the face through involuntary facial expressions and blood flow,” said Prof Ugail.

Humans were known to have a limited repertoire of around 46 facial expressions, known as Facial Action Units (FAUs), he said.

These were used when programming the software for the facial lie detector. The thermal camera focused on the region around the eye because the skin is thinner there, making its temperature changes easier to spot.

The camera is capable of detecting a warming in the face of 0.1C. An early test of the system found that liars could be identified with 70 per cent accuracy.