Inspiration
Project Story: SafeHer AI💡 InspirationThe inspiration for SafeHer AI came from a harsh reality: in moments of extreme danger, a person often doesn't have the time or ability to unlock their phone, find an app, and press an SOS button. We realized that safety tools need to be proactive, not reactive. We wanted to build a "Silent Guardian" that listens and feels when a user is in distress, even if the phone is in their pocket. Our mission was to bridge the gap between human fear and emergency response using Artificial Intelligence.🏗️ How We Built ItWe envisioned an ultra-modern, cyberpunk-style dashboard that feels like a military-grade security tool. The project was built using a Multimodal AI approach:Frontend: Crafted with HTML5 and CSS3 using Glassmorphism and neon aesthetics to create a futuristic "Neural" feel.Audio Intelligence: We integrated the Web Speech API to constantly monitor for distress keywords like "Help Me" or "Bachao".Motion Sensors: Used JavaScript's devicemotion events to detect high-G force shakes, symbolizing a struggle.Backend: A Python Flask neural core that processes risk scores using a simulated ML logic.Real-time Alerts: When triggered, the system initiates a loud 120dB siren and simulates emergency API calls to police and family.🧪 The Logic Behind Risk ScoringTo determine the threat level, we use a probability-based approach. The Risk Score ($R$) can be represented as a function of environmental triggers:$$R = \sum_{i=1}^{n} (w_i \cdot s_i)$$Where:$w_i$ is the weight of the sensor (e.g., Voice = 0.6, Shake = 0.4).$s_i$ is the intensity of the trigger detected.If $R > 75\%$, the CRITICAL SOS protocol is automatically engaged.🚩 Challenges We FacedAudio Latency: One of the biggest hurdles was making the voice recognition work in real-time without lagging the UI. We solved this by using asynchronous JavaScript functions.False Positives: Initially, even a small movement would trigger the SOS. We had to calibrate the $G$-force threshold to ensure it only reacts to violent shaking or struggles.Cross-Browser Permissions: Getting microphone and GPS access to work seamlessly across different browsers required deep diving into browser security policies.📚 What We LearnedThis project was a massive learning curve for Team SafeHer Innovators. We learned how to:Integrate hardware sensors (Mic/Accelerometer) with a web interface.Build a responsive, high-end UI that doesn't compromise on performance.Manage a full-stack Flask application with real-time data logging.Most importantly, we learned that Technology is most powerful when it serves a human cause.
Log in or sign up for Devpost to join the conversation.