🖐 Arabic Sign Language Bidirectional Translator

A bidirectional translator for Arabic Sign Language, converting speech → sign and sign → speech in real-time using your webcam.
This project supports letters, numbers, and words/phrases recognition with live Arabic text output.

---

✨ Features

1️⃣ Speech → Sign

• Type a word or sentence manually or use your voice 🎤.
• Displays animated GIFs for words in Arabic.
• Supported words include: "mom", "Alhamdulillah", "angry", "house", "how are you", and more.
• GIFs are fetched directly from online URLs 🌐.

2️⃣ Sign → Speech

• Real-time hand sign recognition via webcam 🎥.
• Supports three model types:
  • Numbers (0–10)
  • Arabic letters
  • Words/Phrases
• Uses MediaPipe Hands for hand tracking 🖐.
• Recognizes repeated signs and speaks them aloud in Arabic 🔊.
• Displays shaped Arabic text on screen using arabic_reshaper + python-bidi.

3️⃣ Data & Models

• Collected ~500 images per class for letters, numbers, and words.
• Landmarks extracted and saved in pickle files for training.
• Trained Random Forest classifiers for each category.
• Real-time predictions use these pre-trained models.

4️⃣ Controls

• Start/stop the camera 🎥
• Enable/disable speech 🔊
• Toggle between right/left hand ✋
• Clear recognized text 🧹

---

🛠 Dependencies

Python 3.8+ and the following libraries:

pip install customtkinter pillow requests SpeechRecognition opencv-python mediapipe numpy arabic-reshaper python-bidi gTTS pygame scikit-learn

📸 Screenshots

---

Voice → Sign

See how typed or spoken text is converted to animated sign GIFs instantly.

Sign → Speech

Real-time hand sign recognition shows the corresponding Arabic text and speaks it aloud.

---

🚀 How to Use

---

Text/Speech → Sign

1. Go to the "Text → Sign" tab.
2. Type a word and click Show GIF, or click Click to Record to use speech input.
3. The corresponding GIF will appear.

Sign → Speech

1. Go to the "Sign → Speech" tab.
2. Select Numbers, Letters, or Words.
3. The webcam will detect your hand signs.
4. Recognized signs are:
   • Spoken aloud in Arabic 🔊
   • Displayed as shaped Arabic text 🖋
5. Use the Clear Text button 🧹 to reset.
6. Toggle hand using Use Left/Right Hand button ✋

---

💡 Notes

• Audio is cached in audio_cache/ for faster playback 🎧
• Left-hand recognition mirrors coordinates for accuracy ↔️
• Minimum interval prevents repeated audio output ⏱
• Internet is required to display GIFs 🌐
• Maintain equal number of images per class (~500) for better accuracy
• Models can be upgraded to CNN/LSTM for higher accuracy 🤖

---

🔮 Future Improvements

• Expand dataset with more words and phrases 📚
• Implement deep learning models for higher accuracy 🔥
• Add a GUI interface for interactive feedback 🖥
• Support multiple hands and gestures simultaneously ✌️