An American Sign Language Translation System Based on Kinect
In this project, students have attempted to implement a computer-vision based translation system for American Sign Language (ASL). A set of ASL signs, including “Hello”, “How are you”, “Good”, “Thank you”, “Yes”, “What”, “Blue”, “Noise”, and “Tall”, were recorded via a computer program using the Kinect sensor. The Hidden Markov Model (HMM) was used as the machine learning classifier to recognize ASL signs. The HMM classifier was implemented in C# and trained using the recorded data. Unfortunately, the experimental results were disappointing. The accuracy of ASL sign recognition is not acceptable. We conclude that alternative template-based method might be more suitable for the ASL sign recognition. Nevertheless, the project provided a good learning experience for students, and the recorded ASL signs will be valuable for future research.
Sinyuk, Vitaliy; Yokoyama, Sam; Bujdos, Kelsey; and Zhao, Wenbing, "An American Sign Language Translation System Based on Kinect" (2013). Undergraduate Research Posters 2013. 41.