[mythtv] Massively distributed commercial flagging?

[Mike Benoit]

Similar to the P2P commercial flagging thread found here:

http://www.gossamer-threads.com/lists/mythtv/users/84891?
search_string=peer;#84891

I too have been thinking that there has got to be a better way to do
commercial flagging. Especially since its my belief that somewhat poor
reception really wreaks havoc with the current commercial flagging code
(which is expected). 

Lets take for instance 3 machines that all record the same episode of
CSI. 


Machine 1:
Total Program Length: 59:12
Total Commercials: 4
1: Start: 25.56%, Length 60secs
2: Start: 53.20%, Length 182secs
3: Start: 72.98%, Length 120secs
4: Start: 85.64%, Length 90secs

Machine 2:
Total Program Length: 62:34
Total Commercials: 4
1: Start: 22.56%, Length 62secs
2: Start: 51.20%, Length 185secs
3: Start: 71.98%, Length 119secs
4: Start: 78.64%, Length 88secs

Machine 3:
Total Program Length: 57:01
Total Commercials: 4
1: Start: 28.56%, Length 58secs
2: Start: 56.20%, Length 175secs
3: Start: 73.98%, Length 119secs
4: Start: 81.64%, Length 93secs

Each recording will of course be different lengths, and each commercial
flagged at different times. But couldn't all this data be processed
(averaged?) and at least "very good hints" be given to the commercial
flagger? If a central database "in the sky" kept all the commercial
flagging data for each episode that was uploaded to it, and when queried
simply returned the "best guess", could this not cut commercial flagging
time down to just seconds or minutes? Or at the very least make it much
more accurate? Especially if people submit manually created cutlists?

Assuming the central database had the above 3 commercial flagging
results (hopefully it would have many more), when queried it could
return:

Total Program Length: Min: 57:01 Avg: 59:25 Max: 62:34
Total Commercials: 4
1: Min: 22.56% Length: Min: 58secs   
   Avg: 25.56%         Avg: 60
   Max: 28.56%         Max: 62

2: Min: 51.20% Length: Min: 175secs   
   Avg: 53.53%         Avg: 180
   Max: 56.20%         Max: 185

3: ...
4: ...


In theory, based off this information could the commercial flagger not
get away with only scanning 10-25% of the recording, rather then 100%? 
Not only that, but if commercial flagger did scan the entire recording,
it would at least have something to compare against to help eliminate
anomalies like 7minute long commercials, 30seconds apart, etc...

To those more familiar with the commercial flagger, does this at least
sound feasible?

-- 
Mike Benoit <ipso at snappymail.ca>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://mythtv.org/pipermail/mythtv-dev/attachments/20041028/2b506bc3/attachment.pgp