System to Translate English Video or Audio to Equivalent PSL Sentences
No Thumbnail Available
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UMT, Lahore
Abstract
Population of deaf and hearing impaired personnel is increasing day by day and communication gap between hearing impaired and unimpaired is larger than ever. Very few efforts have been done for Pakistan to bridge this gap, this research is intended to facilitate deaf community of Pakistan by proposing grammar based conversion system for English and Urdu speech from a media file to equivalent Pakistan Sign Language sentences. For this, we have used API’s and services like Google Speech Recognition to recognize words spoken in the media file, Google Translation to translate Urdu to English, Machine Learning based python script to punctuate English transcript, Google Natural Language services to Extract Sentences from Transcript and our English to Pakistan Sign Language translation module for conversion to English Sentences to equivalent Pakistan Sign Language Sentences. Our system can recognize audio from multiple video and audio sources which enables us to use this system on vast scale. We gathered data, audio/video sentences and stories, to validate and improve accuracy of our system. With use of well-developed and intelligent services we were able to achieve 99.30% accuracy in case of sentence recognition and 94.25% accuracy in case to story recognition.