[mythtv] New Method for Commercial Detection.

Robert Dunn anakinberrybread at paradise.net.nz
Tue May 2 18:07:11 UTC 2006

Lucas Meijer wrote:
>   if I 
> don't record anything for two weeks,
> chances are good all the known commercials have stopped airing, making 
> the system unable to "relearn" the
> new commercials..
One possible way to reduce this is to run the commercial detection all 
the time, if the PVR is switched on and tuned to a channel.

> One quickly goes to think of an online database of known commercials 
> (probably region specific), but that
> off course would have problems of its own: (deliberate) corruption of 
> the database with incorrect fingerprints,
> being the first one that comes to mind.
This isn't as much of a problem as you would imagine, the reason this 
technique can work is because commercials repeat, where as programs do 
not. Even if someone were to submit keys for the latest episode of a TV 
Show. They wouldn't detect the program unless it aired again before the 
keys 'timed out' and were removed from the database. Over here there are 
very few TV shows which would have the same episode aired within a 
couple of weeks of each other.

>think that
>this method would have to be used in conjunction with other methods to
>determine where the blocks start and end, otherwise it could be very
>cpu intensive to analyze a whole hour long show second by second looking
>for signatures that match the known lists.  In your testing, how have you
>handled this kind of situtation or have you just done analysis to compare
>two clips to see if they contained the same audio.  Another thing that
>will complicate this is the fact that analog broadcasts will not be
>identical every time, so you wouldn't always get a 100% match as you would
>with an exact digital copy.

In my embedded DSP device, I could generate keys and check a smallish database of keys (with no search optimizations at all) in real time. The largest time consuming thing is the checking of a database. The goal for my project I mentioned is to achieve commercial detection in real time, which depending on the number of keys needed I think should be achievable. I dont know much about MythTV (im slowly learning) but is the commercial flagging done in real-time or as a post processing effect?

This method relies on spectral components of the audio, and is highly robust (to quote the research paper) to noise so I dont imagine the analog broadcast would be a problem.



More information about the mythtv-dev mailing list