Extracting almost unintelligible speech[SOLVED]

Hello All,

I use Audacity on two flavors of Linux (Suse and Ubuntu) and I also have it on a Win 7 computer. There are subtle differences but it probably doesn’t matter which one I use unless someone thinks otherwise.

I am hoping someone could suggest a method of attack. I have an audio file that I’m trying to extract some speech from. It was recorded through a closed door and it is very low with the background noise (hiss) usually overwhelming the speech. I’ve had some success applying different combinations of effects: eq., noise reduction, amplify, compress, etc., however, it is pretty much hit and miss. I can’t help but get the feeling that if I did just one more thing or did them in a different order it would greatly improve, but maybe not.

I can hear parts of the speech from time to time so I know it’s there and could probably be made much better.

I was also wondering if 1) could I manually adjust the wave form, & 2) would that matter at all (I was thinking if I could it might help.

Any thoughts would be appreciated.

Our tools all assume the valuable performance is louder than the trash. The minute you reverse that, we’re dead.

We do not offer a free audio forensics service. Please contact an appropriate commercial company.


The obvious answer is to re-record with the door open, or preferably with the microphone in front of the person speaking, but I guess that isn’t an option, in which case, as Koz wrote, an audio forensics service is probably your best bet.

You know that answer was actually helpful. I searched and searched on different topics but couldn’t find what I was looking for probably because I wasn’t using the right terminology. Audio FORENSICS! Duh! Thanks guys.