Automatic Video Subtitling System for Arabic/English Using Speech Recognition
The goal of this project is to develop an application that allows subtitling of English andArabic. The application will provide the user with a full subtitle le ready to be usedby the user. The subtitling will be based on the CMU-Sphinx speech recognition enginewhich is open source GNU software. In this project, the ASR engine will be mostly usedas a black box and the role of the student would be to understand how to integrate theengine with his application.
- Main 6 Steps:
1- extract audio from video ( FFMPEG ) 2- convert audio to 16khz one channel ( FFMPEG ) 3- finding out where each speech pause in the audio ( LIUM ) 4- segmenting the audio at each pause ( FFMPRG ) 5- speech recognize each segment ( CMU SPHINX ) 6- save to an srt file
- Libraries used:
1- LIUM : Detecting segments 2- FFMPEG : Converting file 3- SPHINX : Speech recognition engine 4- AudioSegmentor : Cuts audio to many segments 5- SubtitleWritet : Write the subtile file
- List needed by sphinx to do speech recognition:
1- the decoder source code 2- the language dictionary 3- the acoustic model 4- the language model 5- the test data