AI-Powered Local Wake-Word SOS Activation System

Abstract - Women’s safety applications often fail in emergencies because they rely on manual input, internet connectivity, or GPS availability. In many critical situations, victims may not be able to use their phones, and poor network coverage decreases the reliability of existing SOS apps. To tackle these challenges, SOS Activation System offers an AI-powered Offline Women Safety Alert System that works without internet and with minimal user interaction. The system uses a TensorFlow Lite keyword-spotting model that constantly listens for phrases like “Help me” to trigger an emergency alert offline. Once activated, it sends distress signals to nearby devices through Wi-Fi Direct, Bluetooth Mesh, and ultrasonic high-frequency transmissions. This ensures message delivery even in areas with no network. Hands-free triggers such as shake detection, fall detection, and proximity activation further improve usability when the victim cannot reach the device. The SOS Activation system also automatically records audio, video, sensor data, and the last known location as evidence, storing it securely for transmission once connectivity is restored. This multi-layered approach offers a reliable, practical, and innovative solution for improving women’s safety in both online and offline situations. Key Words: Women's Safety, Offline SOS, TensorFlow Lite, Bluetooth Mesh, Wi-Fi Direct, Emergency Communication.