x
anonymous 10 June 2010

Hi folks,

I was looking around the Internet searching for discussions related to this topic (bleeding in audio recording) in many forums.

I see people have a lot of approaches and different opinions on how to deal with this issue, or even on if one has to worry about it at all.

Right now, I am studying a set of methods and techniques called blind source separation, and thinking about applying it to this problem of instrument bleeding in audio recording. That is, the goal is to have a computational tool that can help audio engineers to address this problem by letting them input an audio track with a "main instrument" and some bleeding of others instruments and get a track only with the main instrument as the output.

I wonder if you think that such a tool would be really useful in your projects.

Any opinion will be appreciated.

Thank you!

Comments

anonymous Thu, 06/10/2010 - 19:03

Thank you.

Yup, sure.

Roughly speaking, it is a set of computational methods which consist in estimating some original signals that went through an unknown mixing process, using only the mixed signal(s). That is, both the mixing process and the original signals are unkwnown, and little (if any) assumptions are made about them (this is why the word "blind" is used).

It is used very successfully in many problems on fields like telecommunications and biomedical signal processing.

MadMax Thu, 06/10/2010 - 22:18

For some types of music, this would probably be very useful. The opposite engineering would be EXTREMELY useful as a noise reduction system...

Still... for performance based tracking.. I wouldn't destroy what I have worked so hard to achieve by making the bleed work for me in a constructed plan of capturing the sound energy as a position of distance.

Ultimately, the ideal blend of the performance is two signals which are balanced to the sound of the room/enclosure, where all instrumentation is captured equally, and the dynamics of the performers in the environment are what bring out the intricacies of the art of musicianship.

Now, IF you can blind source the frequency spectrum, regardless of phase relationship, fast enough to pull the phase relationship to 180' of both positive and negative excursions, of any two signals... you could potentially create a system whereby you could essentially create holes in the separation.

Or, you could shove an omni below a figure 8, and create a mid-side, when you track it.

But creating M/S encoding/decoding in a plug would be sweet, no doubt. You might actually have a shot a re-creating a decca tree if yer' able to handle all 4 phase relationships... or you can build a decca tree and capture it when tracking.... so.... it comes back to that same ol' argument...

If your stuff works, and works well, is reasonably priced, and you can actually stand behind the product for more than 10 years... you could be golden.

x

User login