Lie Detection Based on Human Facial Gestures

Abstract

Facial behaviour of deception has developed as an imperative subject in observation research. In this paper, we assess the facial conduct of deception focused around some of the signals identified in the literature. To start with,we present a short study of deception cues in facial behaviour. At that point, we run a pilot study, with one-to-one interview with balanced situation, under controlled and exploratory conditions. Emulating the standard practice, the test has been intended to join three stages. Phase one: help session, in which the baselines are made for every member. Phase two: session, in which the members are obliged to answer addresses on two points - one truthfully and one erroneously. Phase three: reporting toward one self session, in which the members are obliged to report the ground truth for stage two. Once the information of pictures of facial conduct was gathered, it was broke down and contrasted with the facial behaviour prompts identified in literature. For measurable examination, a non-parametric sign-test was chosen. The results demonstrated that there is potential in applying facial behaviour as the signs in predicting deception. At last, an examination on future work present more unpretentious facial actions analysis is presented. In this thesis, We have worked on facial action points and expression analysis. We will take a shot at the strategies to decrease the lapse in Lie Prediction. We have likewise taken mind that utilizing Simple Computer Webcams ,one can have the capacity to utilize our System. In this dissertation, We focus on building an algorithm which can predict a Lie and a dataset which consists of different facial actions that are observed on the face during a Lie. We have investigated that micro expressions cannot be controlled for every emotions. So, we have approached towards various techniques to collect the action units and expressions that can be helpful for Lie observation on the face.