[Mythtv-soc] Commercial Flagging Update July 7 2006

[Randy Robertson]

This week I've continued work reworking the commercial detection
framework to allow additional methods in a plug and play fashion, and
started investigating new detection methods. By this time next week I
hope to have completed a simple audio compression detection method,
and will be doing extensive testing looking at output produced on a
variety of nuvs I've got laying around.  I will examine how reliable
this detection method is, combined with existing methods.  After
looking into accessing VBI data, it seems that inconsistencies in
accessing this data across the multitude of v4l devices supported with
myth would make any implementation only useful to a small cross
section of users. Although I'd like to be convinced otherwise, if
anyone suggests a specific direction that seems the most likely to
benefit the largest amount of users.  Time permitting (or if I get
stuck with audio analysis, and decide to do this first) I'll begin
with my method for storing and analyzing past skiplist data for
similar shows as described in the my original outline.  I think this
method will probably be one of the most useful methods for the largest
amount of people, as much TV content is episodic, and detection might
benefit greatly from prior knowledge of previous commercial cut
points.