ahmad-bensalah/SignBridge

SignBridge is a real-time Tunisian Sign Language translation app that uses a TuniSL BiLSTM recognition model and a Tunisian Arabic TTS model to turn signed gestures into speech. It captures webcam input, extracts MediaPipe landmarks, predicts words/phrases, and speaks them aloud for smoother deaf-hearing communication. # 🇹🇳 SignBridge: Bridging the Gap in Tunisian Communication **SignBridge** is a comprehensive accessibility platform designed to break communication barriers between the deaf and hearing communities in Tunisia. By leveraging state-of-the-art Deep Learning models, the platform provides seamless bidirectional translation between **Tunisian Dialect (Darija)** and **Tunisian Sign Language (TSL)**. --- ## 🚀 Project Vision In Tunisia, the deaf community often faces significant hurdles due to the lack of specialized tools that understand the nuances of the local dialect and native sign language. TuniSign AI aims to provide a "Conversation Bridge" that is: - **Culturally Aware**: Trained on Tunisian speech and sign datasets. - **Real-time**: Capable of live sign recognition and instant speech synthesis. - **Hyper-Visual**: Utilizing neon-skeleton avatars for better clarity and engagement. --- ## 🔍 How it Works: Step-by-Step ### 🖐️ Sign to Text (Hearing the Signer) 1. **Capture**: The app uses the camera to record the user's hand signs. 2. **Skeleton Tracking**: Mediapipe "sees" the hands and maps out 21 key points (knuckles, fingertips). 3. **Data Analysis**: These points are converted into numerical coordinates and angles. 4. **Time Learning**: The AI watches how these numbers change over a 16-frame sequence. 5. **Translation**: The system outputs a Tunisian word (e.g., "3aslema"). ### 🤟 Text to Sign (Showing the Sign Language) 1. **Input**: A word is typed or received in Tunisian Darija. 2. **Database Lookup**: The app finds the exact hand-movement data for that word. 3. **Skeleton Rendering**: The "Neon Avatar" uses those points to move its limbs. 4. **Visual Synthesis**: The movements are smoothed out to look like a real person signing. 5. **Display**: The deaf user sees the glowing skeleton sign the message visually. ### 🗣️ Text to Speech (The App's Voice) 1. **Text Normalization**: Tunisian text is cleaned so the AI knows how to pronounce loanwords. 2. ** …