home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.speech:270 comp.dsp:2597
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!sgiblab!darwin.sura.net!paladin.american.edu!news.univie.ac.at!hp4at!mcsun!julienas!babbage!imag!alissali
- From: alissali@imag.fr (Mamoun ALISSALI)
- Newsgroups: comp.speech,comp.dsp
- Subject: Re: disguizing identity of talker
- Message-ID: <43769@imag.imag.fr>
- Date: 16 Nov 92 11:11:34 GMT
- References: <1992Nov12.145247.24996@mnemosyne.cs.du.edu>
- Followup-To: comp.speech
- Organization: IMAG Institute, University of Grenoble, France
- Lines: 23
-
- In article <1992Nov12.145247.24996@mnemosyne.cs.du.edu> sliberm@nyx.cs.du.edu (Sergio Liberman) writes:
- >
- >What kind of algorithms exist that are capable of
- >disguizing the voice of a person?
- >
- [...]
- Are there other methods?
-
- If I got the question right, I'd suggest you to read about synthesis by
- analysis methods. One that is becoming more and more used is OLA (OverLap and
- Add) with its variants (PSOLA, TD-PSOLA). The idea is to analyse registered
- sound segments, extract some parameters, modify appropriate value (like pitch,
- which might answer your question), then reconstruct the signal with the new
- parameters.
- If this is what you want I could give some pointers to the subject.
-
- Mamoun.
- >
- >Sergio
- >---
- >sliberm@nyx.cs.du.edu
-
-
-