Training Computers To Recognise Microexpressions: A University of Oulu Research

If you have had the chance of watching the short lived American TV show “Lie to Me”, you might know a thing or two about microexpressions and their importance in lie detection tests. The show was based on the work done by American psychologist, Paul Ekman on facial microexpressions that reveal deception in humans. Ekman found out in his research that most people cannot pick up these microexpressions without training but if researchers led by Xiaobai Li at University of Oulu in Finland have their way then they could design computer software that is much more efficient than these select individuals at picking these subtle expressions. Before delving into this research we would like to talk about microexpressions in the following paragraph.

Microexpressions are subtle and quick facial changes that appear and disappear on people’s faces in matter of microseconds when they are trying to hide their real emotions. For example, when your boss reprimands you for your work, you may, for a quarter of second display emotions related to disapproval or hatred on your face before you put up your best straight face. While the concept of microexpressions is heavily debated among psychologists across the world, there is a booming industry that is devoted to training law enforcement officials and security officials to pick up on those microexpressions.

The team at University of Oulu in Finland took the existing concept of machine vision that is being used to recognise human expressions and improved it to recognise microexpressions. The first task was to build a database of microexpressions that would be compared against the current footage. Since microexpressions are generated by individuals under stress, they employed 20 people and asked them to watch a series of videos designed to invoke strong emotions among them. For each time they failed to hide their feelings they were asked to fill a lengthy and boring questionnaire. This punishment made sure that they worked hard to hide their actual emotions. The researchers videotaped the reactions of the test subjects on a high-speed camera at 100 frames per second.

Microexpressions

Next, the task of capturing microexpressions in other people which was tackled by breaking down footage of a reaction into frames and comparing each frame with another with help of a computer. Since the changes were very subtle, the researchers had to magnify the muscle movements of certain parts of the face to help the machine recognise these changes. Under test conditions the machine was able to outperform humans but it still has a long way to go before being used in the real world.

For more information on this research, head over to the #-Link-Snipped-# on the Cornell University Library website and its coverage on #-Link-Snipped-#.

Replies

You are reading an archived discussion.

Related Posts

I AM STUDYING MICROCONTROLLER 8051 SUBJECT, WHILE DOING PROGRAM OF TIMER OPERATION, I HAVE FOUND SOME PROBLEM IN THE MODE 2 OF TIMER OPERATION IN WHICH CONTENTS PRESENT IN THx...
Indian smartphone manufacturer Micromax has put up its new smartphone, the Canvas Amaze for sale. A few days after surfacing on the company’s official website, the Micromax Canvas Amaze went...
how to choose any specific filed , like engineering , doctor , Vakil etc
what are the ways to increasing efficiency air conditioner?
Besides working on cell phones and being a telecommunications services company, Huawei has also been working on bringing new technologies to market. In the first of such work, Huawei has...