Embedded System for Societal Violence Detection and Intervention

In developing countries, harassment and social violence pose significant challenges and making life difficult. To prevent such violence, an end to end solution can provide a way of indirect communication to the victim. This work presents an embedded system with a triggering mechanism activated by hand gestures. When a predefined gesture indicating distress is detected, the system captures images of all faces in the frame and sends them via email to a predetermined address for intervention. The Haar cascade and MediaPipe method has been used for face and hand gesture detection respectively. The method implemented on a Raspberry Pi module is found to trigger alerts up to 8 feet away in dim lighting with 100% accuracy in crowded environments. The average latency for triggering and subsequent actions is measured at 6.3 seconds. In summary, our integration of security cameras with embedded technology can provide an accessible approach to harassment detection and intervention.