I have a few wma-files all about 2h long each.

I was going around with a microphone interviewing people. I did not turn the recording off in order to not miss anything. Now the problem is filtering the interviews.

It is two-fold:

  1. I want to find the areas in which there is a conversation between me and a person.
  2. I want to exclude background noise as well as walking noise created while rubbing on the microphone's foam.
  • I think this is less of a programming question and more of a "I need a tool question". – obrok Jul 23 '11 at 21:17
  • @obrok I've edited the question to reflect that. Programmatic solutions are one of many possible solutions to this problem. – Warrior Bob Jul 24 '11 at 19:30

To extract the conversations, load up your recordings into a sound editor like Audacity. You can use any editor that shows you a waveform, Audacity is just an example and it's open-source (read: free) so it's a good starting point.

Look at the waveform. Chances are, the parts with conversations in them are going to have the highest sustained volume levels, since someone is speaking directly into the mic - so if you see a part that has a higher average level to it, give it a listen. You can trim these parts out manually and save them as new files or new tracks within the same editor project; however you want.

To reduce noise, the best solution I've personally ever seen used is Izotope RX, although it's fairly pricey. Audacity has a noise-reduction tool that I've never had much luck with, but you might give it a try.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy