[hts-users:02072] Vector Size conflict
- Subject: [hts-users:02072] Vector Size conflict
- From: Girish Malkarnenkar <girish1m@xxxxxxxxx>
- Date: Tue, 7 Jul 2009 12:39:09 +0200
- Delivered-to: hts-users@xxxxxxxxxxxxxxx
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:reply-to:date:message-id :subject:from:to:content-type; bh=S51k/TUwulvksQcuC0CXSLLbHzN+JE9RBiGbnWa9dDQ=; b=YhKjCtQLpepIzPigmTRXCUMM+j9eRxxVOOD+QT7PdF0ZW8FOkmNmJpx8Mk1akt86BQ ObPETkaKDMxg2UYNkJCTQdg7rx4L+sQoZVnOM4T9c8i4eynwhpYVo4qpAcHoL5X6POVM q9Dve6E3Uw5e+nz0w03uXVu8L43p1pJ4dIdCw=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:reply-to:date:message-id:subject:from:to:content-type; b=GpW0qAJ0KkvGAsAE9TP/OKp2IUR27PjWGJ7hK+JlNEt99vz6epjlYrbb5Kec5PfgSK hRvAyPkPL07qqupV4HrqMoBKsQgqgPE/ehFCIT0nsrHD6wOoPoloGDNY+m80UfafW/bP bPGWVGbdGgCiYE2O4i/mlQ6qfUM/ulFP88+3g=
Dear Sir/Madam,
I am working on facial parameter synthesis. For this purpose I replaced the .MGC files created by my facial vector binaries and created empty dummy lf0 files. I also changed the MGCORDER to MGCORDER = 194 # order of MGC analysis in the data/makefile and also in # feature order in the
config.pm. When I ran the training.pl I got an error that:
Can't take log of 0 at scripts/Training.pl line 1280.
Since the problem was with
# convert GV pdf -> GV mmf
if ($useGV) {
conv_gvpdf2mmf();
}
$useGV = 0; # turn on GV
Now my first doubt is whether GV is needed for generation of MGC parameters using 1mix or hts_engine?
Secondly after I bypassed this step I got an error during the computation of variances that
ERROR [+2050] CheckData: Vector size in /home/hp/germanexps/data/cmp/german1.cmp[78] is incompatible with hmm /home/hp/germanexps/proto/ver1/state-5_stream-4_mgc-585_lf0-3.prt[588]
FATAL ERROR - Terminating program /cygdrive/c/Users/hp/Desktop/PROJECTPHASE1/FINAL/HTK-3.4/htk/HTKTools/HCompV
Error in /cygdrive/c/Users/hp/Desktop/PROJECTPHASE1/FINAL/HTK-3.4/htk/HTKTools/HCompV -A -C /home/hp/germanexps/configs/trn.cnf -D -T 1 -S /home/hp/germanexps/data/scp/train.scp -M /home/hp/germanexps/models/qst001/ver1/cmp -o /home/hp/germanexps/models/qst001/ver1/cmp/init.mmf /home/hp/germanexps/proto/ver1/state-5_stream-4_mgc-585_lf0-3.prt
Now as far as I can make out the .cmp was created using the wrong number of values per vector (order of 25, *3 for delta and delta-delta = 75, I have 3*195 = 585. I think I need to modify the part where the cmps are compiled so that "Vector size in /home/hp/germanexps/data/cmp/german1.cmp" is "[585]. However I checked the part in the /data/makefile where the .cmp files are created and I couldn't find any line of code that placed the value of 78 (3*25+3) in the cmp files including the header. Can anyone please help me and tell me what I need to modify in order to change the .cmp file info of vector size?
Yours sincerely
Girish Malkarnenkar
- Follow-Ups
-
- [hts-users:02073] Re: Vector Size conflict, Matt Shannon