Indian Sign Language Recognition

Overview

This project focuses on recognizing Indian Sign Language (ISL) gestures using video input. It leverages computer vision to extract hand landmarks from video frames and utilizes a Long Short-Term Memory (LSTM) model to classify gestures. The system is designed to aid communication for individuals with hearing and speech impairments.

Features

• Real-time Video Processing: Captures video input to recognize ISL gestures.
• Landmark Extraction: Uses MediaPipe or OpenPose to extract hand landmarks.
• Gesture Recognition: LSTM-based model for sequence classification.
• Scalability: Supports training with an extended set of ISL gestures.
• User-Friendly Interface: Can be integrated into applications for accessibility solutions.

Workflow

1. Video Capture: The system records a video of hand movements.
2. Landmark Detection: Extracts hand landmarks (X, Y, Z coordinates) from each frame.
3. Feature Processing: Normalizes and processes the extracted coordinates.
4. Sequence Classification: Uses an LSTM model to classify gestures.
5. Prediction Output: Displays or converts recognized gestures into text or speech.

Technologies Used

• Python (OpenCV, NumPy, Pandas)
• MediaPipe (for hand landmark detection)
• TensorFlow/Keras (for LSTM model training and prediction)
• Matplotlib & Seaborn (for data visualization)

Installation

1. Clone the repository:

   git clone https://github.com/UnbeatableBann/Indian Sign Language.git
   cd Indian Sign Language

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   python main.py

Dataset

• The model is trained on a dataset of ISL gestures.
• Each gesture video is converted into a sequence of landmark coordinates.
• The dataset includes common ISL gestures like greetings, alphabets, and numbers.

Model Training

1. Preprocess Data: Extract landmarks from videos and store them in CSV or NumPy arrays.
2. Train LSTM Model:

   python train.py

3. Evaluate Model: Assess accuracy using test datasets.

Future Enhancements

• Expand gesture vocabulary for better recognition.
• Implement a real-time translation system.
• Improve accuracy with attention-based neural networks.
• Develop a mobile application for on-the-go recognition.

Contributors

• Your Name (@yourgithub)

License

This project is licensed under the MIT License.

Acknowledgments

• Inspired by research on sign language recognition.
• Thanks to the open-source community for resources and tools.