pakistani-sign-language-recognition-system
We developed a system that would translate Pakistan Sign Language to text and audio form in real time. The system is limited to recognition of 15 pre-specified greeting sentences. System takes input from a Kinect Camera device as a person practices signs in front of it. The system generates the sentences on the screen being practiced in the sign. System uses depth information right from the Kinect device and computes the time-series of a sign being practiced. System is trained on features extracted from 15 sentences using Random Forest Machine Algorithm. System evaluates the input sign and on model established after training and displays the word against the sign practiced on the screen as suggested by model. System can be extended to commercial scale based on the schema proposed by the project. The intended subjects of this technology are the people who suffer from any kind of hearing loss and knows Pakistan Sign Language and those who want to communicate to a mute person and do not know Pakistan Sign Language.
