Hi,
I'm currently in the bootstrap phase of compiling a database of impulse
responses of common end-user microphones (specifically those on
headsets). With this, it is in theory possible to correct some of the
aspects of the microphones, processing the signal to be closer to that
of a studio microphone. The motivation is simple; no matter how
expensive and perfect my headset is, everything still sounds like crap
if the user at the other end has a $15 headset.
The current idea is detailed at
http://mumble.sourceforge.net/MicrophoneDatabase (along with a few
issues we could use help with).
Assuming the owners of the acoustics lab agrees, I plan to publish the
recordings under the BSD license along with some simple matlab/octave
code to create a inverse filter (and the C code to apply it at runtime).
If anyone here has some suggestions or ideas (or want to send us
headsets to test), I'd love to hear it :)
Best Regards,
Thorvald
