SUBPAL: A Device for Reading Aloud Subtitles from Television and Cinema

The primary focus of this paper is accessibility barriers for visually impaired people and people with dyslexia. Due to their disability a segment of these people have limited accessibility to the subtitle content presented on television and in the cinema. It is utterly important to be able to participate in such social and cultural events, but if the material presented is in a non familiar language they are unable to understand it. The problem primarily arises from non English speaking countries where dubbing is not facilitated such as in Scandinavian countries and the Netherlands. A solution to this problem is SubPal, a text to speech device which can be connected to the television or to a video camera. The subtitle content in the presented video stream is read aloud through a multilingual speech synthesizer. Hence the solution is applicable for television and in the cinema, in several countries. The solution comprises three major modules: The sampling of the analogue video signal into a binary image of the subtitles. The optical character recognition which converts the binary image of the subtitles into a characters that can be recognized by a computer. And finally a speech synthesizer that reads the decoded subtitles aloud. The system is quantified and a method for sampling the video signal is proposed and verified. Requirements to the optical character recognition algorithm is discussed, and parallel studies on such algorithm is referred to. The speech synthesizer is discussed in the context of user and technical requirements, and a best candidate synthesizer is evaluated. The conclusion implies potential for further studies towards a prototype.