Intelligent Facial and Emotion Recognition System for the Predictive Assessment of Socially Dangerous Behavior
Objective. This article addresses the automation of monitoring in public spaces to facilitate the early detection of signs of socially dangerous behavior. Method. The system is built upon a multi-level computer vision pipeline that integrates state-of-the-art open-source neural network models: InsightFace for face detection and embedding generation; DeepFace and FER as independent emotion classifiers; and MediaPipe Face Mesh for the geometric analysis of facial expressions. Result. An intelligent face and emotion recognition system has been developed to predict the risk of conflict in crowded public areas. A key feature of this solution is an adaptive buffering and frame-selection mechanism, which isolates the most informative images for subsequent analysis. This significantly enhances the system's robustness in real-world conditions, such as varying lighting, diverse viewing angles, and partial occlusions. A crucial aspect of the design is the implementation of "privacy by design" principles: data processing is localized, biometric features are represented as anonymized embeddings, and a multi-level, role-based access control model is applied to sensitive data. Conclusion. The developed system demonstrates practical utility as a predictive analytics tool for security services, enabling a shift from reactive surveillance to the proactive identification of potentially dangerous situations.
Authors
- Andrey Gazizov (ORCID: https://orcid.org/0000-0002-1441-4357)
- Elena Aleksandrovna Revyakina (ORCID: https://orcid.org/0000-0003-1577-2671)
Institutions
- Don State Technical University (RU)
Publication Details
- Journal
- Herald of Dagestan State Technical University Technical Sciences
- Published
- 2026-11-12
- DOI
- https://doi.org/10.21822/2073-6185-2026-53-3-119-133
- Primary Topic
- Face recognition and analysis
- Type
- article
- Field-Weighted Citation Impact
- 0.00