From: Graham Mitchell <graham@gr...> - 2006-08-25 12:47:09
> Separating individual voices:
Cool article, but don't get your fingers crossed. It's only using a neural
net at the moment, which is clever but not really generally reproducible.
And musical instruments tend to have a lot higher range of frequencies than
the human voice merely speaking, so it really really would compound the
Give it 10 years and you could maybe record a three-piece band in a club and
then use this algorithm to split the recording back out into three separate
tracks, each one sounding like a MIDI-produced version of the original.
I hate to rain on anyone's parade, but I didn't want this article to falsely
get hopes up for the non-programmers on the list.
Graham Mitchell - computer science teacher, Leander High School
"Christians and their opponents again and again expect that some new
discovery will either turn matters of faith into matters of knowledge or
else reduce them to patent absurdities. But it has never happened. What we
believe always remains intellectually possible; it never becomes
intellectually compulsive. I have an idea that when this ceases to be so,
the world will be ending."
-- C.S. Lewis, from "Religion and Rocketry"