[mythtv-users] Re: Torrentocracy Patch against Mythtv-cvs

Brad Templeton brad+myth at templetons.com
Thu Mar 3 21:29:11 UTC 2005


On Thu, Mar 03, 2005 at 01:59:04PM -0800, Aaron Stewart wrote:
> Not necessarily so.. Depending on reception quality (I know that my
> channel 3 comes in worse than most), there may be more or less latency in
> blank frame detection.  Ditto to positions that are set by user when
> flagging.
> 
You don't need supreme accuracy on this level.  If you have an error
level, you just buffer a bit, at the risk of showing the user a second
or two of something boring like a commercial.  They can always pick
up the remote and correct manually if it gets too far off, this is
what we do with automatic commercial detect.

But the truth is you don't use just one approach.  You have a lot of
ways of looking at it.   Since most users will have NTP, you have
of course the exact time for people watching the same broadcast.
Then you have any clues you could pick out of the show, from the
existing detection of commercial stuff (blank frames, scene transitions,
logo on-off etc.)  You have VBI information, in particular the
closed captioning.    I have not had the chance to look at different
streams from different broadcasters, but I strongly suspect that
the closed captioning is synced to the video once and exactly once,
and if you have a frame with a caption on it, it's the exact same
frame (or very close) everywhere the show is broadcast.

(No good on stuff without captions of course.)

All of these things could quickly lead you to learn that most users
skipped over a given section of video, and thus identify it as boring.

Not that this is an easy project, this is one of the more complex
things on the large mythtv wishlist.  But I think it's doable.


More information about the mythtv-users mailing list