We've sent a verification link by email
Didn't receive the email? Check your Spam folder, it may have been caught by a filter. If you still don't see it, you can resend the verification email.
Started March 25th, 2015 · 5 replies · Latest reply by Headphaze 9 years, 7 months ago
I have an audio recording of a mobile phone, but it is very poorly understood speaking persons. On the recording same time recorded music on the radio, which cancels interview. The radio is a song that I as a separate file. How can I remove background music from audio recording? For example, applications Audacity. Thank you.
Phase inversion comes to mind (but there are some problems with mono and stereo signals - you may need to convert your music and audio files both to mono, perhaps). Check out this video, and you may want to do some research along these lines:
https://www.youtube.com/watch?v=yUJMvWSAIz8
You're right, however...
Phase inversion only works properly if the two waveforms match exactly and are 'in phase' (you're right about matching the channel number), so if you have a recording of a track playing from a speaker and then the music file itself, it will probably fail because the audio signals are entirely out of phase due to reflections and the recording mediums.
I'm afraid this is a very tricky situation, for which you're going to need some more sophisticated tools like iZotope RX2 if you want to remove the majority of the music. Or just record it again without the background music, which would be by far the easiest option (if available).
You can always try phase inversion of course, it might have a small effect at attenuating higher frequencies that have a discernible pitch. Realistically, there's no way to completely remove the background music without completely destroying the dialogue along with it. It's a shame you only have audacity at your disposal, because it's extremely limited for this kind of task.
Personally, in this particular scenario, this is the chain that I would use to attenuate the music whilst keeping the dialogue relatively intact:
> De-reverb
> Phase inversion
> De-Noise
> Spectral repair (algorithmic)
> Gate with noise floor just above -infdB
In theory, if you are patient and expert enough, you can use direct FFT spectral repair much in same way as photoshop, and "paint" out any unwanted audio from a signal, but I can assure you this will take hours upon hours of work to complete, so it's not practical to do this kind of stuff.
Hope this helped
Thank you very much for posts. I really helped, although the result is not ideal. Because the record is very poor quality, noise, and 8-bit quality, succeeded only partially amplify some places. Where the music was too strong, I was unable to suppress it. Thank you very much for your help.