/American-Sign-Language-Real-Time-Translation

Primary LanguagePythonMIT LicenseMIT

Real-time American Sign Language (ASL) Fingerspelling Translation

Introduction

American Sign Language (ASL) is the predominant sign language for the Deaf community in the United States. Because most signs involve movement and facial expressions, this project aims to only recognize ASL fingerspelling.

Below are signs that this program will recognize.

ASL Signs

Requirements

  • Numpy (1.13.0)
  • Scikit-Learn (0.18.1)
  • OpenCV (3.0)
  • Scipy (0.19.0)
  • Python3

Installation

  • git clone https://github.com/GarrettBeatty/American-Sign-Language-Real-Time-Translation.git
  • pip3 install -r requirements.txt

Running

  • python3 webcam.py

How it Works

  • Extract features from the training images using the SIFT algorithm.
  • Create a dictionary of visual words using the KMeans algorithm.
  • Create a Bag of Words (BoW) model for each image.
  • Feed the BoW model into a machine learning algorithm. (This program using a Support Vector Machine.)

Results

This program is still a work in progress. Future updates include potentially using a neural network to improve results.

Author

  • Garrett Beatty

License

See License

More Info

https://en.wikipedia.org/wiki/American_Sign_Language# American-Sign-Language-Real-Time-Translation