[hts-users:03203] A bug in HFB.c
- Subject: [hts-users:03203] A bug in HFB.c
- From: "Heiga ZEN (Byung Ha CHUN)" <heigazen@xxxxxxxxxx>
- Date: Thu, 15 Mar 2012 21:43:35 +0000
- Delivered-to: hts-users@xxxxxxxxxxxxxxx
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20120113; h=mime-version:date:message-id:subject:from:to:content-type :x-system-of-record; bh=fIQba84zl0/x7iYQa16M/RzPObtVJa7UwcoNZY3FqtM=; b=QA2JKAZxtlWZRcUzs2fP6XaC66auzYxhzw3KdcBmMpxzvx929+6qzjh3ZaNzbba+e9 nieYnTkCitMPvOpvTJKOSFz/DfgGzAigMGA8GN7KcTGwshG6dcvZG1Zr7cV6V3/w/Mt+ mkFRfUlMn1H1nWFeSjwzMfsH+Tjb4cOiLSDs9S2JEk5+6PGTIQx6D8bQVNnxiD2L3ogN gntCICyXq7OIWghxGn0gQq2eAVf8ywYthmkM+IRb2gkmMPAVFV9yb48Xk1wf5f7NQp4w 2LIKKIOF2KqGOAfL7RsZ2D8RFCy8QOIR7sYT03dSBNwy21WMHlb/wsSRQUWpEngWmXw6 h6SQ==
Hi,
I found a bug in HFB.c; two model reestimation with mixture component
level alignments fails if models have multiple streams. Here is a
patch code.
--- HFB.c
+++ HFB.c
@@ -2058,7 +2058,7 @@
norm = LZERO;
for (mx=1; mx<=M; mx++) {
if (alCompLevel) {
- al_ste = al_hmm->svec[j].info->pdf+1;
+ al_ste = al_hmm->svec[j].info->pdf+s;
al_sti = al_ste->info;
if (al_sti->nMix != M)
HError(999,"Cannot align at the component level
if number of components is different!");
Regards,
Heiga
--
Heiga ZEN (in Japanese)
Byung Ha CHUN (in Korean)
<heigazen@xxxxxxxxxx>
- Follow-Ups
-
- [hts-users:03204] Re: A bug in HFB.c, Keiichiro Oura