Patrol System for Aggression Detection
Description
The article presents the development and validation of **TacDeAPred**, an intelligent and modular system designed for the **real-time detection and prediction of aggression risk** during patrol operations conducted by security personnel. The system integrates multiple sensors (camera, microphone, heart rate monitor, and GPS) with advanced artificial intelligence techniques, including LSTM neural networks optimized by genetic algorithms to forecast emotions and heart rate, and a fuzzy logic model with 12 inference rules to assess risk levels. Implemented on a portable Raspberry Pi-based platform and validated with 20,000 real-world labeled records, TacDeAPred achieved over 94% accuracy and sub-second response times, demonstrating its effectiveness in anticipating aggressive behavior and enhancing operational safety for agents in complex environments.
Files
Steps to reproduce
The project requires a Python platform with Jupyter Notebook, and all required libraries are already integrated into the repository and the source code