AI Polygraph studies your face and voice for signs you're being deceitful

Looking for tiny facial tics from opponents has been used by poker players for decades as a way to spot a bluff, but now Chinese scientists are using the same technique to develop a lie detector.

They have created an online polygraph that uses facial recognition to study minute changes in facial expression, skin colour, temperature and heart rate - all of which could reveal a lie.

The machine is also fitted with artificial intelligence to analyse visual and voice data to look for such changes that indicate someone is not being truthful.

Scientists in China have developed a polygraph that combines artificial intelligence with facial recognition. It assesses changes in facial expression, skin colour, temperature as well as voice to spot if someone is telling the truth or not. A picture of an earlier prototype AI facial recognition system in use is shown above

Scientists in China have developed a polygraph that combines artificial intelligence with facial recognition. It assesses changes in facial expression, skin colour, temperature as well as voice to spot if someone is telling the truth or not. A picture of an earlier prototype AI facial recognition system in use is shown above

Traditional lie detectors rely upon detecting changes in a person's pulse, breathing rates and skin conductivity, but these all require someone to be fixed up to sensors.

By contrast, the facial polygraph can be used without contact, raising the prospect of individuals having their truthfulness tested without them realising it.

Cameras for the device look for tiny changes in facial expression, changes in colour which indicate changes in heart rate, and fluctuations in temperature as blood flow changes.

The researchers behind the system claim it would be particularly useful in overcoming the problems faced when conducting polygraphs on people who have been trained to defeat them.

Researchers said their system overcomes many of the problems of traditional polygraph tests (reconstruction pictured) and can be used without physical contact. By using cameras and voice detection, it means people can have their truthfulness tested without being aware they are being assessed by the system

Researchers said their system overcomes many of the problems of traditional polygraph tests (reconstruction pictured) and can be used without physical contact. By using cameras and voice detection, it means people can have their truthfulness tested without being aware they are being assessed by the system

HOW TO SPOT A LIAR 

Using videos from criminal court cases, researchers at the University of Michigan claim to have developed machine-learning software that can tell if someone is being deceitful by looking for specific 'tells.'

From their results, these are the most common ways to spot a liar: 

Scowling/grimacing: People who lied were more likely to scowl or grimace using their whole face. 

This was in 30 per cent of lying videos versus 10 per cent of truthful ones.

Eye contact: Liars tended to look directly at the questioner in 70 per cent of deceptive clips compared to 60 per cent of truthful.

Gesticulating wildly: Gesturing with both hands was also spotted among the people who withheld the truth - 40 per cent compared to 25 per cent. 

Using 'um' and 'er': People who were lying in the clips tended to speaking with more vocal fills such as 'um.' 

Speaking confidently: Liars also had a tendency to make statements with more confidence and certainty than people who were telling the truth. 

Using 'he' or 'she' in place of 'I': To seemingly distance themselves from the action, liars were found to use words such as 'he' or 'she,' rather than 'I' or 'we' when withholding the truth.   

It has been developed by researchers at the Chongqing Institute of Green and Intelligent Technology at the Chinese Academy of Sciences, the Shanghai Jiao Tong University and a company called Cloudwalk.

Dr Zhou Xi, chief executive of Cloudwalk and a scientist at the Chongquing Institute who led the project, said: 'Traditional polygraphs usually use a contact sensor to detect blood pressure, pause and muscle activities of the targets, but sometimes it doesn't work on people who have received professional training.

'The advantage of our new polygraph is the non-contact detector, which will lower the vigilance of targets.'

'It can work in both daylight and dark environment, and offer more accurate test results,'

Dr Zhou said the new polygraph hinges around artificial intelligence technology that uses five different types of 'recognition modules' to assess if someone is being truthful or not.

It uses intelligent decision making to recognise changes in facial expression, voice, semantic structure, physical posture and morphology.

About Unknown

Unknown
"Mình là Phương Nguyễn, thâm niên 4 năm kinh nghiệm thiết kế website và làm marketing, tuy nhiên kể từ 1 năm trở lại đây mình không còn làm marketing nữa, và chỉ tập trung vào viết plugin và giao diện cho Wordpress, nếu các bạn thấy bài viết hay thì hãy chia sẻ cho những người khác cùng tham khảo, còn nếu muốn thiết kế website hoặc sửa web hay đặt một plugin có chức năng đặc biệt, hãy liên hệ ngay tới Phương"
Recommended Posts × +

0 nhận xét:

Đăng nhận xét