🌟 Minki – Bridging Silence with Technology
🧠 Inspiration
In a world full of communication, millions of deaf and mute individuals face daily challenges expressing themselves, especially in public places. We wanted to build something affordable, portable, and powerful—a solution that could serve as a real-time interpreter for sign language, making everyday interactions seamless and inclusive.
💡 What it does
Minki is a small, smart device that uses a camera module, Raspberry Pi, display screen, and audio output to interpret American Sign Language (ASL) gestures in real time and convert them into spoken and visual words.
Key Features:
- 🖐️ Real-time hand gesture recognition using MediaPipe and a custom ASL-trained ML model
- 📷 Captures signs via the camera and sends coordinates to the ML model
- 🔊 Converts recognized gestures into voice output using text-to-speech
- 🖥️ Simultaneously displays the interpreted text on the device screen
- 🔌 Fully portable and designed for public interaction (banks, stores, hospitals, etc.)
🛠️ How we built it
- Hardware: Raspberry Pi 4, Pi Camera Module, Speaker, Touchscreen Display
Software:
- MediaPipe for gesture landmark detection
- Custom TensorFlow/Keras model trained on ASL dataset
- Python for backend logic
- Pyttsx3 for text-to-speech conversion
- Lightweight Flask server for modular testing
Deployment: The full model runs on the Raspberry Pi without needing internet—making it fast and offline-friendly.
🧩 Challenges we ran into
- Achieving high FPS with limited Pi processing power
- Gesture overlap (e.g., letters with similar signs)
- Power management and device compactness
- Fine-tuning the model to handle different hand sizes, lighting conditions, and backgrounds
🚀 Accomplishments that we're proud of
- Successfully interpreting ASL gestures into speech with 85%+ accuracy
- Fully self-contained device—no cloud dependency
- Super intuitive UI and minimal latency
- Designed with accessibility and cost-efficiency in mind
📚 What we learned
- How to deploy ML models efficiently on edge devices
- The intricacies of real-time gesture detection
- Importance of optimizing UX for accessibility
- Working with hardware-software integration
🔮 What’s next
- 🔁 Add support for two-way communication (speech to sign via an animated hand model)
- 🌍 Multi-language and regional sign language support
- 🤝 Partnering with NGOs and accessibility groups
- 📱 Companion mobile app for updates and customization
🧑💻 Built with
Python, TensorFlow, MediaPipe, OpenCV, Raspberry Pi, Pyttsx3, ASL Datasets, Hardware Prototyping
✨ Team
A group of passionate undergrads dedicated to making the world more inclusive—starting with Minki 💙
Built With
- cameramodule
- gru
- lstm
- mediapipe
- numpy
- opencv
- python
- raspberry-pi
- rnn
- scikit
Log in or sign up for Devpost to join the conversation.