On 4/8/2015 3:00 PM, John Williamson wrote:
> What are you trying to get, though?

The voice and only the voice at the end.

> The only voices which sound
> recoverable in that track are already clear

Clear to you (or any human ear) but not to a Speaker Identification or 
Voice Recognition software. I intend to verify that the voice belongs to 
the gentleman in the picture, by using audio clips known to be from him.

This is the software that I plan to use:

     http://alize.univ-avignon.fr/

-Ramon

On 08/04/2015 20:32, Ramon F Herrera wrote:
>
> Dear audio/DSP experts:
>
> Please go to my *Google Drive* and click once on the *AT&T* folder.
>
>       http://goo.gl/Jm4bZ7
>
> Watch the TV commercial (MP4) and listen to the background music.
>
> Since my objective is to apply some *Voice Identification* process to
> the audio, as a preliminary step, I would like to remove the music.
>
> I have already identified the song as "You're My Best Friend" by Queen.
>
> Being a newbie, I am only familiar with:
>
>   • Adobe Audition
>   • iZotope RX4
>   • Audacity
>
> but I don't believe they have such feature. I gather that you profs use
> something called DAW?
>
Those or other similar programmes running on a computer form a Digital 
Audio Workstation.

> Note: I found a free app which performs subtractions, but has been
> abandoned and it is really meant for signals which have very close content.
>
> http://www.libinst.com/Audio%20DiffMaker.htm
>
> How common is the "difference" operation? If it works as advertised, it
> should be extremely useful, right?
>
>
Depending on what you're trying to extract, you could try a karaoke 
filter, which will give you the mix less anything positioned central in 
the sound field, then subtract the result from the untreated track. 
Audacity has at least one available, though you might have to search the 
Web for a plugin.

What are you trying to get, though? The only voices which sound 
recoverable in that track are already clear, and it sounds as if the 
final voice may not even be speaking English. Anything more would be in 
the field of forensic audio analysis and extraction, which is a much 
harder problem, often needing dedicated hardware and software.


-- 
Tciao for Now!

John.

Dear audio/DSP experts:

Please go to my *Google Drive* and click once on the *AT&T* folder.

      http://goo.gl/Jm4bZ7

Watch the TV commercial (MP4) and listen to the background music.

Since my objective is to apply some *Voice Identification* process to 
the audio, as a preliminary step, I would like to remove the music.

I have already identified the song as "You're My Best Friend" by Queen.

Being a newbie, I am only familiar with:

  • Adobe Audition
  • iZotope RX4
  • Audacity

but I don't believe they have such feature. I gather that you profs use 
something called DAW?

Note: I found a free app which performs subtractions, but has been 
abandoned and it is really meant for signals which have very close content.

http://www.libinst.com/Audio%20DiffMaker.htm

How common is the "difference" operation? If it works as advertised, it 
should be extremely useful, right?

TIA,

-RFH