Hi,
I've been trying to build a regression tree for speaker adaptation. I am
using HTS 2.1.1.
The command used to build a regression tree is
/usr/local/HTS-2.1.1/bin//HHEd -A -B -C /home/speech/HTS_tamil_train_mal_adapt/configs/trn.cnf
-D -T 1 -p -i -H /home/speech/HTS_tamil_train_mal_adapt/models/qst001/ver1/cmp/re_clustered.mmf
-M /home/speech/HTS_tamil_train_mal_adapt/models/qst001/ver1/cmp/regTrees
/home/speech/HTS_tamil_train_mal_adapt/edfiles/qst001/ver1/cmp/reg.hed
/home/speech/HTS_tamil_train_mal_adapt/data/lists/full.list
The problem was that splitting of nodes did not finish. It seemed to
be in a endless loop, and the score stayed the same. So the HHEd command cannot
resulted in a error.
The log file is as below :
HHEd
126874/126874 Models Loaded [7 states max, 2 mixes max]
// load stats file
LS /home/speech/HTS_tamil_train_mal_adapt/stats/qst001/ver1/cmp.stats.re-clustered
Loading state occupation stats
Stats loaded for 126874 models
Mean Occupation Count = 3.491070
// construct regression class tree
RC 32 reg
Building regression tree with 32 terminals (4 streams)
Creating regression class tree with ident reg.tree and baseclass reg.base
Splitting Node 1, score 1.000000e+10
(Stream splitting)
Splitting Node 3, score 1.000000e+10
(Stream splitting)
Splitting Node 5, score 1.000000e+10
(Stream splitting)
Splitting Node 7, score 1.000000e+10
(MSD splitting)
Splitting Node 10, score 1.006489e+10
(Stream=3, vSize=1)
Splitting Node 4, score 1.000000e+10
(MSD splitting)
Splitting Node 2, score 3.957550e+06
(Stream=1, vSize=75)
Splitting Node 16, score 1.597189e+06
(Stream=1, vSize=75)
Splitting Node 18, score 7.206715e+05
(Stream=1, vSize=75)
---
---
---
Splitting Node -18811, score 6.065758e+03
(Stream=3, vSize=1)
Splitting Node -18809, score 6.065758e+03
(Stream=3, vSize=1)
Splitting Node -18807, score 6.065758e+03
(Stream=3, vSize=1)
Splitting Node -18805, score 6.065758e+03
(Stream=3, vSize=1)
Splitting Node -18803, score 6.065758e+03
(Stream=3, vSize=1)
Error in /usr/local/HTS-2.1.1/bin//HHEd -A -B -C
/home/speech/HTS_tamil_train_mal_adapt/configs/trn.cnf -D -T 1 -p
-i -H /home/speech/HTS_tamil_train_mal_adapt/models/qst001/ver1/cmp/re_clustered.mmf
-M /home/speech/HTS_tamil_train_mal_adapt/models/qst001/ver1/cmp/regTrees
/home/speech/HTS_tamil_train_mal_adapt/edfiles/qst001/ver1/cmp/reg.hed
/home/speech/HTS_tamil_train_mal_adapt/data/lists/full.list
Can anyone help me how to solve this endless splitting node situation?
I have used three speakers data for training. The text used is the
same for all the three speakers.
Thank you,
regards
Lilly