Help Regarding Source Seperation

Hi guys, after going through various articles on BSS, ICA, CASA etc etc in
vain I finally come here to ask you guys for help, the problem is as
follows:

Given a noise free recording of a particular person's voice I need to
extract the voice charactaristics of that person's voice and then use
those
charactaristics to filter his/her voice from a single channel recording in
which multiple persons voice and background noise may be present.

Any solution to this problem or any thing that will help me get a good
start on how to solve it will be most welcome, thanks in advance.