Early Warning System Based On The Analysis of Facial Microexpressions For The Detection of Suspicious Behavior In Security Environments

Public safety in public spaces has become an increasing concern, leading to the adoption of advanced technological solutions, such as artificial intelligence, to accurately identify potential threats. This study presents the development of an early warning system developed by the University of the Armed Forces ESPE, which uses the YOLOv8 model, selected for its high accuracy and effectiveness in real-time face and Convolutional Neural Networks capable of analyzing subtle facial micro-expressions. During system testing, it achieved an accuracy rate of 81%, delivering positive results and demonstrating high performance. Additionally, a complementary analysis based on psychomorphology has been incorporated, exploring the relationship between facial features and behaviors, such as anxiety expressions or evasive postures. The system successfully addresses technical challenges, including lighting fluctuations and partial obstructions, which often limit other models, ensuring its applicability in real-world scenarios. This technological advancement is designed for implementation in airports, transportation hubs, and large-scale events, such as concerts or stadiums, providing a valuable enhancement to existing surveillance systems. With its ability to detect potentially dangerous behaviors, this development highlights the role of artificial intelligence as a key tool for threat prevention, public safety improvement, and societal protection.