Gabe
Newbie level 1
- Joined
- Jun 3, 2003
- Messages
- 1
- Helped
- 0
- Reputation
- 0
- Reaction score
- 0
- Trophy points
- 1,281
- Location
- All Over the World
- Activity points
- 11
I am working on a project to create a decent de-vocalizing algorithm. A devocalizing algorithm removes voices from stereo audio for purposes such as Karaoke. The problem with this is that current techniques leave behind flanging effects and other disturbing artifacts.
The most popular (simple) way to create a devocalizer is to take a two channel wav file and invert both Right and Left Channels and Mix them with the original Opposite Channel. Because a Mono signal is identical and evident in both channels this eliminates most of the voice. However, it also sometimes eliminates drums or instruments also in the "mono" channel.
A small improvement is made when the lower frequencies are filtered using a LP butterworth filter at 300Hz. (and re-added after the devocalizer).
Anyway, does anyone have any better suggestions for a better implementation of this or can someone direct me to better info :?:
thanks - Gabe
The most popular (simple) way to create a devocalizer is to take a two channel wav file and invert both Right and Left Channels and Mix them with the original Opposite Channel. Because a Mono signal is identical and evident in both channels this eliminates most of the voice. However, it also sometimes eliminates drums or instruments also in the "mono" channel.
Code:
New RightChannel = Inverted Left Channel + Right Channel;
New LeftChannel = Inverted Right Channel + Left Channel;A small improvement is made when the lower frequencies are filtered using a LP butterworth filter at 300Hz. (and re-added after the devocalizer).
Anyway, does anyone have any better suggestions for a better implementation of this or can someone direct me to better info :?:
thanks - Gabe
