SignVoice AI

Bridging Voices Beyond Silence

---

**GitHub Link - https://github.com/imsinghaditya07/SignVoice-AI

👥 Team BYTE BREAKER

*Anand Kumar Jha (Team Leader | (Backend Developer)

*Aditya Singh (AI & System Architect | UI/UX Designer)

*Debashrita Mandal (Integration & Deployment)

*Aaryan Lal Das (ML Model Train | Data & Training)

---

📌 Overview

SignVoice AI is a real-time, AI-powered bi-directional communication system designed to bridge the gap between deaf/mute individuals and people who rely on spoken language.

The platform enables seamless communication by converting:

• ✋ Sign Language → Text
• ⌨️ Text → Sign Language (visual form)

This system focuses on accessibility, simplicity, and real-world usability, making communication inclusive for everyone.

---

🎯 Problem Statement

• Millions of deaf individuals struggle to communicate daily
• Most people do not understand sign language
• Lack of affordable and accessible assistive tools
• Communication barriers in hospitals, schools, and public services

---

💡 Our Solution

SignVoice AI provides a two-way AI communication bridge that allows:

• Deaf users to express themselves using gestures
• Hearing users to communicate using text

👉 Without requiring either person to learn the other’s language.

---

⚙️ Key Features

• 🔄 Bi-directional communication
• ⚡ Real-time gesture recognition
• 🧠 AI-powered prediction
• 📷 Webcam-based sign detection
• 🖼️ Sign language visualization
• 📱 Responsive and simple UI
• 🌍 Accessible for rural users
• 🚨 Emergency gesture detection (HELP)

---

🧠 Tech Stack

👨‍💻 Frontend

• React.js / Next.js
• Tailwind CSS
• Framer Motion

🤖 Backend

• Python
• Flask / FastAPI

🧪 AI & Machine Learning

• OpenCV
• MediaPipe (Hand Tracking)
• Scikit-learn (Random Forest Model)

🔊 Processing

• Text Processing Engine
• Dataset Mapping (Sign Images / GIFs)

---

🧩 Functional Architecture

🔄 Bi-Directional System

SIGN LANGUAGE → TEXT

Webcam → MediaPipe → Landmark Extraction → Feature Processing → ML Model → Prediction → Text Output


TEXT → SIGN LANGUAGE

Text Input → Text Processing → Sign Mapping Engine → Dataset (Images/GIFs) → Visual Output

---

⚙️ System Components

• Input Layer

  • Webcam (gesture input)
  • Text input

• Processing Layer

  • MediaPipe hand tracking
  • Feature extraction

• AI Layer

  • Machine learning model (Random Forest)

• Output Layer

  • Text display
  • Sign language visuals

---

🚀 How It Works

🟢 Sign → Text

1. Webcam captures hand gesture
2. MediaPipe detects 21 landmarks
3. Features are extracted
4. ML model predicts gesture
5. Output displayed as text

---

🔵 Text → Sign

1. User enters text
2. Text is processed
3. System maps text to sign visuals
4. Output displayed as sign language

---

📊 Dataset

• Custom dataset created using MediaPipe landmarks
• Multiple gesture samples collected
• Labels include common words like: HELLO, YES, NO, HELP

---

📈 Performance

• Real-time processing
• Lightweight model
• Works on standard devices
• Accuracy: ~85–95% (depends on conditions)

---

🌍 Impact

• Enables communication for deaf individuals
• Useful in hospitals, schools, and public spaces
• Promotes inclusivity and accessibility

---

🔮 Future Scope

• Sentence-level recognition (LSTM / Deep Learning)
• Multi-language support
• Mobile application
• 3D avatar-based sign display
• Offline AI model

---

🏆 Innovation Highlights

• Real-time bi-directional communication
• Low-cost and scalable solution
• Human-centered AI design
• Practical real-world application

---

🎯 Vision

“To create a world where communication is not limited by ability.”

---

📜 License

This project is developed for educational and research purposes.

---

❤️ Acknowledgements

• MediaPipe
• OpenCV
• Scikit-learn
• WHO for data references
• Open-source community

---

⭐ Final Note

SignVoice AI is not just a project — It is a bridge that connects silence to expression.

If you like this project, give it a ⭐ and support inclusive technology!