HydrogenAudio

Lossy Audio Compression => MP3 => MP3 - Tech => Topic started by: sreejith on 2010-07-26 07:28:00

Title: Audio Source Seperation
Post by: sreejith on 2010-07-26 07:28:00
Hi,

Can i  separate sound of each instrument and vocal from an audio input? i need to separate the harmonics spatially and need to increase quality of each source using java or c++ code. Is it possible? and How?

Thank you.
Title: Audio Source Seperation
Post by: DVDdoug on 2010-07-26 20:14:07
If I understand your question, the answer is, No! "You can't un-fry and egg or un-bake a cake, and you can't un-mix a song"

See this post (http://www.hydrogenaudio.org/forums/index.php?showtopic=82464)[/color].

With a stereo recording, you can subtract the left and rigtht channels to remove the "center channel" vocals (and everything else in the center), and of course you can isolate the left & right channels.


BTW - You don't have to write code to subtract the channels.  "Vocal Removal" is built-into most audio editors, and if not, any audio editor will allow you to invert one channel and mix.
Title: Audio Source Seperation
Post by: sreejith on 2010-07-27 06:13:35
What is the meaning  of spatially separating harmonics from an audio? i think that spatial separation results the separation of instrument sounds.
Title: Audio Source Seperation
Post by: mjb2006 on 2010-07-27 07:03:22
What is the meaning  of spatially separating harmonics from an audio? i think that spatial separation results the separation of instrument sounds.


You're the one who asked how to spatially separate harmonics, and now you're asking what spatial separation of harmonics actually is? Have you been assigned a task that you don't even understand the meaning of? I think you need to go back to whoever asked you to do this and get a clearer understanding of what they want. And instead of promising to do it for them, then desperately inquiring on message boards, you might have to consider just telling them "no, it's not possible."

Did you even read the previous reply?
Title: Audio Source Seperation
Post by: Martel on 2010-07-27 07:21:41
If you have a multichannel recording where the position of microphones is known, there are mathematic methods of "focusing" on a particular source. I think this is mostly used to extract a particular voice from strong ambient noise (or other voices). It's kind of similar to how a human is able to focus on a particular remote conversation during a party (spatial information provided visually).

I can't provide more information as this was a Ph.D. thesis of one of my schoolmates, not mine.
Title: Audio Source Seperation
Post by: sreejith on 2010-07-27 07:28:27
Thank you.