ASL Recognition with Text-to-Speech

This project is an implementation of an American Sign Language (ASL) recognition system using computer vision and machine learning. The system recognizes ASL letters from a live webcam feed and speaks the recognized letters out loud using text-to-speech functionality.

This project is based on a YouTube tutorial by Computer Vision Eng, but I added my own twist by integrating text-to-speech functionality to make the system more interactive and user-friendly.

---

Features

• Real-Time ASL Recognition: Recognizes ASL letters from a live webcam feed.
• Text-to-Speech Integration: Speaks the recognized letters out loud using the pyttsx3 library.
• Customizable Dataset: Allows users to collect their own ASL data for training.
• Machine Learning Model: Uses a Random Forest Classifier for letter recognition.
• Hand Landmark Detection: Utilizes MediaPipe Hands for detecting hand landmarks.

---

How It Works

1. Data Collection:

   • Use collect_images.py to collect ASL data for each letter.
   • The script captures sequences of frames for each letter and saves them as .jpg files.

2. Dataset Creation:

   • Use create_dataset.py to process the collected data.
   • Extracts hand landmarks from the frames and saves the processed data in a data.pickle file.

3. Model Training:

   • Use train_classifier.py to train a Random Forest Classifier on the processed dataset.
   • The trained model is saved as model.p.

4. Real-Time Inference:

   • Use inference_classifier.py to recognize ASL letters in real-time from a webcam feed.
   • The recognized letters are spoken out loud using the pyttsx3 text-to-speech library.