The first thing to do is to find a way to make a screenshot with your app (sourcecode for that can be found in the forum) This is probably easier to do when the TV stream is played with a video stream player embedded in your project.
You will need to get hold of the stream for each station (often m3u8 types of files or starting with rtmp:// or rtsp://)
I expect that you will need to take screenshots about every second.
If you play the TV streams with an external player (other app), then I doubt that Android will allow you to capture that screen (because such configurations might be construed as spyware). I am planning to experiment with that myself (putting the screenshot routine in a service) because I want to make a subtitle reader for my blind clients.
As OCR solution I suggest you take a look at the OCR project(s) by Johan Schoemann. OCR uses a lot op processor power. I am not sure whether the OCR will be able to keep up with the refresh of new subtitle lines. The whole OCR process should be faster than analysing one TV screenshot about every 2 seconds...