Edge Artificial Intelligence

12A Lightweight Lie Detector for Resource Constraint Devices

Mayur Navin Sharma, Rohan Deep Kujur, Khushi Tulsian, Tejaswini J. Chavan* and Atharva Manish

School of Technology Management and Engineering SVKM’s NMIMS Navi-Mumbai, Maharashtra, India

Abstract

Over the course of technological advancement, numerous machines and software have been developed to detect lies. Presently, lie detection methodologies, such as polygraph testing, rely on physiological signals and thermal imaging to detect subtle changes indicative of deception. Despite their widespread use, these approaches primarily gauge physiological discomfort rather than directly detecting lies, and they are impeded by significant expense. In this study, we propose leveraging deep learning models to enhance lie detection accuracy by analyzing facial expressions, iris movement, and body posture. The integration of these techniques presents a promising avenue for augmenting lie detection capabilities, potentially surpassing the constraints inherent in traditional methodologies.

General Terms: Lie detection; deep learning; image processing

Keywords: Facial micro-expressions, vision system, neural network, body posture, head movement, iris movement

12.1 Introduction

Deception detection, a critical endeavor in various societal contexts, has historically relied heavily on subjective human judgment, leading to inconsistent outcomes and potential biases. “The polygraph and Lie Detection” [7], concludes that polygraph technique ...