[mythtv-users] Commercial Flagging Problem

[Chris Pinkham]

> > I am running MythTV .16 using a PVR-350.  Commercial flagging was very
> > accurate for a while, but now is almost random.  The areas identified as
> > commercials often are regular program segments, and actual commercials are
> > detected infrequently.  What could have caused this deterioration? The
> > image quality is quite good, so I don't think that is the problem.
> 
> I coul image that if the detection was rather easy the companies would
> take actions to make it harder to detect. If no one watches the
> commercials anymore who is going to pay the TV station?
> 
> So the stations have a significant interrest in making it very hard to
> detect commercial breaks.

It is definitely an art not a science, and to make matters worse, every
capture card is different when capturing analogue TV.  So, for
instance on one user's card #1 a black/blank screen is all values 0-10
but on another card in the same computer, it could be 60-70.  I tried to
code around some things like this in current CVS, but there are still going
to be some differences that I'm not accounting for.  This is one reason you
could have 2 people recording the same show and it works great for one person
but not-so-great for the other person.  Add to that the fact that like you
say the people making the content and stations don't want us skipping
commercials and it makes it practically impossible to get it working
perfectly.  One of the things that I noticed that started breaking
commercial flagging using the blank-frame code in 0.16 was the fact that
CBS started putting their logo up during the blank frames.  Since the old
code didn't know about the logo, it would detect the frame as being
non-blank and it would realy mess up the commercial detection.  In current
CVS, I try to work around that and it seems to be doing much better.
There is also a new "ALL" method which tries to use all available info
from the recording to find commercials.  The ALL method is extensible so
I'll be adding more things to it in the future to help better detect
commercials and eliminate false positives.  I've got code started in
my source tree to try using aspect ratio changes to help detect when
commercials start and stop.  There are also a few other detection ideas
on my TODO to add that will help the ALL method be more accurate.

-- 

Chris