LIE DETECTION USING FACIAL MICRO-EXPRESSION FEATURES WITH MACHINE LEARNING

The development of artificial intelligence and computer vision opened the door to the development of lie detection using facial micro-expressions where detecting a lie using the computer is more useful than the human eye, many authors show the comparison between human detection and machine learning detection they found that artificial intelligence is better than human, despite the important of artificial intelligent in lie detection there is a limitation on using it like blurry and noise that found in the video also the limitation of the real dataset, therefore this paper treated the blurry and noise by using the preprocessing method, then two type of filters and compare between them, for the limitation two types of dataset were used one for training and other for testing.Sobel filter was more useful than canny filter, and the mouth and nose were also more useful in the lie detection than the eye.The recall, precision, and F1 scores of using the Sobel filter with mouth and nose are 0.998, 0.994, and 0.996 respectively for the truth case and 0.981, 1, and 0.990 for the lie case.