Document Types
Paper Presentation
School Code
N/A
School Name
St. Edward School, General Trias City, Cavite
Abstract/Executive Summary
This study explores the feasibility of using low-resolution cameras as a means of detecting facial movements for lie detection. Micro-expressions, however, are difficult to detect by the human eye due to their short duration and low intensity, thus the research explores the possibility of extracting micro-expressions from phone or web cameras that have low resolution and framerate. The collected videos are the processed using time series processing, to obtain both facial data points extracted from facial landmark detection models, as well as image generation from the obtained datapoints to produce a face structure. The classification mainly focuses on the use of common machine learning algorithms, to detect facial movement patterns, in the hopes of classifying people telling truths or lies. The tests ultimately proved to have a low accuracy in classification, but the results show that the methodology may contribute to other domains, such as in person identification, as well as possible recommendations for future works.
Keywords
Facial Landmarks, Image Transform, Machine Learning, Time Series.
Research Theme (for Paper Presentation and Poster Presentation submissions only)
Computer and Software Technology, and Robotics (CSR)
Initial Consent for Publication
yes
Feasibility of Using Phone and Web Cameras to Detect Micro-Expressions for Lie Detection
This study explores the feasibility of using low-resolution cameras as a means of detecting facial movements for lie detection. Micro-expressions, however, are difficult to detect by the human eye due to their short duration and low intensity, thus the research explores the possibility of extracting micro-expressions from phone or web cameras that have low resolution and framerate. The collected videos are the processed using time series processing, to obtain both facial data points extracted from facial landmark detection models, as well as image generation from the obtained datapoints to produce a face structure. The classification mainly focuses on the use of common machine learning algorithms, to detect facial movement patterns, in the hopes of classifying people telling truths or lies. The tests ultimately proved to have a low accuracy in classification, but the results show that the methodology may contribute to other domains, such as in person identification, as well as possible recommendations for future works.
