23 Oct
2011
23 Oct
'11
11:42 p.m.
On 23 October 2011 at 17:38, Fons Adriaensen <fons(a)linuxaudio.org> wrote:
On Sun, Oct 23, 2011 at 08:13:24AM -0700, Kevin
Cosgrove wrote:
I never thought much about requirements, as I wasn't thinking of
creating one.
The subject line says it all. I'm looking
for a Linux solution to
watermarking audio. Thanks...
I've never seen any. The amount of work to develop such a thing
would depend on your exact requirements. Important parameters
are: * Which operations should leave the watermark intact.
If this
includes lossy encoding/decoding things get more difficult.
Maybe version 0.5 would work with lossless and version 2.X would
work with lossy?
* The minium duration of a fragment that allows the
mark to be
detected.
However long is easiest I suppose. 1 sec? 10 sec? Repeated in
spots through the file X times, where 1 < X < (length in minutes)?
* How strong the 'provability' should be.
Maybe version 0.5 is 10% chance of proving, and version 2.X would be
better?
* Operational requirements. For example, should it be
possible
to verify the mark without having the information necessary
to remove it ?
Some operations would be:
- Mark
- Un-mark
- Re-mark
- Remove
- Verify signature
- Verify content
In a later version maybe the tool could be able to use some of the
above tools, in cooperation with other sound file tools, to preserve
a watermark across sound file format conversions, between lossless
forms and crossing between lossy/lossless domains.
For version 0.01 even just being able to attach an md5 sum of an
audio file along with a gpg signature would be cool.
I use a subset of GNUPG's capability. Maybe you can see that in
my above requirements? ;-)
Is that a useful start on a start?
Thanks....
--
Kevin