A person in everyday life badly understands cheating. Studies show that, on average, only in 54 cases, untrained subjects can correctly determine whether other participants in the experiment are honest with them. Nevertheless, lies give different signs - it is not only a change in the physiological indicators of the body (rapid heartbeat or sweating), but also external characteristics. For example, even small changes in the facial expression can reflect the emotions that a person wants to hide.
Modern machine learning systems have a great potential for deception recognition. In many respects this is due to the fact that the computer is trained on a large data set, and this allows him to notice small details that a person simply does not pay attention to. Therefore, researchers from the University of Maryland and Dartmouth College created a special algorithm-polygraph DARE.
At the heart of DARE's work are several existing and pre-trained classifiers and models. For the training of the algorithm, a database was also used, containing 121 videos with a staged trial. The system tracked changes in facial expressions, voice and speech. The principle of the program is shown in the picture below.