Dynamic hand gesture recognition using kinect
Loading...
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Management and Technology Lahore
Abstract
With a population of 70 million deaf people in world [1], Pakistan has approximately 0.3 million [2] deaf people residing on its land. There have been several researches done in developed countries to come with a mechanism to give voice to the deaf community. American Sign Language, British Sign language has been standardized for American and British deaf community consequently. Other countries have developed their own sign language. In Pakistan, deaf community uses Pakistan Sign Language (PSL) as a mode of communication. To overcome the gap between signers and non-signers and to give voice to deaf community, we have proposed an automated solution which uses Microsoft's sensor device, Kinect-1. Our system uses skeleton and depth frame data provided by Kinect. Depth data is used to recognize the hand shape. Other features are computed from skeleton data. Gestures are computed by a hierarchical neural network which is trained using back-propagation method. First network computes the hand Shape. The second network recognizes the gesture performed taking the recognized hand shape and other features as input. The system outputs the recognized gesture in the form of text. The system works with an accuracy of 87.5%.
Description
Supervised by:Nabeel Sabir Baloch Khan
Keywords
Deaf people, British Sign language, BS Thesis