[hts-users:02019] Re: Using parent transforms during speaker adaptation
- Subject: [hts-users:02019] Re: Using parent transforms during speaker adaptation
- From: "Hui LIANG (Tienshyang)" <tshlmail-hts@xxxxxxxxx>
- Date: Tue, 16 Jun 2009 02:35:40 -0700 (PDT)
- Delivered-to: hts-users@xxxxxxxxxxxxxxx
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=yahoo.com; s=s1024; t=1245144941; bh=4zpqqcMICNjs8PutUemE7xzXcednTzs66TdrQy0L0S0=; h=Message-ID:X-YMail-OSG:Received:X-Mailer:Date:From:Reply-To:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=Wb126ftoxnBraLEYerVnS6ZsV52UsgKdIjxDNoDQE+TetTeXh+cIzHNLtpQCrjoiGQ6oytVeWgyEaU5eWTNeQ9FwyYBCFxbH9AHM7Nbfmk00ExH8fhUKt8QqwVkMdvI81KVwHO+EaUQ9MX+GABJ9ajk/Mo+DOPMK4Wly/BOH1fk=
- Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:X-YMail-OSG:Received:X-Mailer:Date:From:Reply-To:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=rResie/7UaCwtcPPytOovGdrSMXv50gfCY/kR77OgpSqq/PMQAVs9LpHKpwW2odsxBCU/fIX162zoS79zUjiXhtLEyZuNbGexn3F66B0juL0o0wQY083i/0VNeZvPp7CNv45KJumRtLDr7nYXF3PBx6zR3jsFCFMOnPDq3ZQ2Q8=;
By the way, I am not sure if the parent (loaded by -E & -W) and input (loaded by -J & -Y) transforms have to share the same regresstion tree and base classes.
--- On Mon, 15/6/09, Hui LIANG (Tienshyang) <tshlmail-hts@xxxxxxxxx> wrote:
> From: Hui LIANG (Tienshyang) <tshlmail-hts@xxxxxxxxx>
> Subject: Using parent transforms during speaker adaptation
> To: hts-users@xxxxxxxxxxxxxxx
> Date: Monday, 15 June, 2009, 6:36 PM
> Hello, everyone,
> I am trying to introduce a set of parent transforms into
> speaker adaptation. I notice that HERest has two options, -E
> and -W, relavant to parent transforms.
> Suppose I have already had an HERest command which works
> well for speaker adaptation without parent transforms.
> Besides, I have parent regression trees, parent base classes
> and parent xforms in the folder "parentdir". Is it enough
> that I simply add "-E ./parentdir cmp.cmllr -W ./parentdir
> dur.cmllr -E ./parentdir -W ./parentdir" into the HERest
> command so as to do speaker adaptation with parent
> transforms? Do I have to modify any configuration files or
> to do anything else?
> Currently, an error message seems to imply that HERest
> didn't find the parent regression trees and the parent base
> Thanks a lot!
> Hui LIANG