Hi All,

I’m exploring Sox for the purpose of reducing background noise from
recordings of people speaking. For my project this will be required for a
large number of files so being able to automate this task via the command
line is a huge plus and it seems that Sox accommodates this.

However, I’ve compared some initial experiments of what Sox’s Noisered can
do to Audacity’s Noise Reduction effect and so far it seems that Audacity
is perhaps a bit more effective. That said, I still need to play around
with Sox more before I can draw any conclusions.

I’ve looked online and in the mailing list archives and haven’t been able
to find any documentation/explanations of what exactly Sox’s Noisered
algorithm is doing….does anyone have an understanding of this, or have a
reference that they could please refer me to? I’m assuming that it’s some
version of spectral subtraction of course, but the devil is in the details
and it would be great to be able to understand what Sox is doing beyond
just a high-level description.

Any help would be greatly appreciated!

Thank you. Best,
Michael