A polygraph is an instrument that records a number of physiological responses and is often used to test whether a person is lying. To be accurate, "lie detection" is not a measure of "lie" itself, but a change in physiological parameters caused by psychological stimulation. And now, there is another high-tech way to test lies.
The data Lab of Tselia, a machine learning company, has recently developed a camera algorithm, which detects lies based on face and face signals. The accuracy rate has been 75%. According to reports, this is called Fraudoscope AI camera system equipped with a high-definition camera, always pay attention to by the facial feature detection object, and the system software is responsible for some of the subtle changes in facial features, such as dilated pupils, hemifacial spasm were described and analysed.
The system is not very different from the traditional lie detector. At the beginning of the test, the detected object will be asked to imagine what kind of gold medal it has just won. When the testers answer the fictitious question, the system will detect and detect some changes of the face details when the lie detector is lying.
At present, the challenge for the system is that although computers can better capture some details of facial details, computers can not yet understand what these subtle changes of the face mean exactly, and the system is still worse in understanding human emotion. The company says it will use machine learning algorithms to make the system better understand human behavior in the future.
The company said that the accuracy rate of 75% is basically the same as that of lie detector. Though it looks not very high, it still has many potential applications to detect lies with AI camera system. Perhaps in the future in the shop to install a camera, can be "aware" of some illegal motive.