get_vtln_warps.sh
3.7 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
#!/bin/bash
# Copyright 2014 Daniel Povey
# Apache 2.0
#
# This script takes a data directory and a directory computed by
# ./train_lvtln_model.sh, and it computes per-utterance warp-factors utt2warp. It
# expects vad.scp to exist in the data directory. Note: like
# train_lvtln_model.sh, it uses features of the speaker-id type, i.e. double
# delta features with sliding window cepstral mean normalization.
# Begin configuration.
stage=-1
config=
cmd=run.pl
logdet_scale=0.0
subsample=5 # We use every 5th frame by default; this is more
# CPU-efficient.
nj=4
cleanup=true
num_gselect=25
num_iters=5 # number of iters of transform estimation
# End configuration.
echo "$0 $@" # Print the command line for logging
[ -f path.sh ] && . ./path.sh;
. parse_options.sh || exit 1;
if [ $# != 3 ]; then
echo "Usage: $0 <data-dir> <vtln-dir> <exp-dir>"
echo "e.g.: $0 data/train_novtln exp/vtln exp/train_warps"
echo "where <vtln-dir> is produced by train_lvtln_model.sh"
echo "Output is <exp-dir>/utt2warp"
echo "main options (for others, see top of script file)"
echo " --cmd (utils/run.pl|utils/queue.pl <queue opts>) # how to run jobs."
echo " --nj <num-jobs> # number of jobs to use (default 4)"
echo " --config <config-file> # config containing options"
echo " --stage <stage> # stage to do partial re-run from."
exit 1;
fi
data=$1
vtlndir=$2
dir=$3
for f in $data/feats.scp $data/spk2utt $vtlndir/final.lvtln $vtlndir/final.dubm $vtlndir/final.ali_dubm; do
[ ! -f $f ] && echo "train_deltas.sh: no such file $f" && exit 1;
done
if [ -f $data/utt2warp ]; then
echo "$0: source data directory $data appears to already have VTLN.";
exit 1;
fi
mkdir -p $dir/log
echo $nj > $dir/num_jobs
sdata=$data/split$nj;
split_data.sh $data $nj || exit 1;
cmvn_sliding_opts="--norm-vars=false --center=true --cmn-window=300"
# don't change $cmvn_sliding_opts, it should probably match the
# options used in ../sid/train_diag_ubm.sh and ./train_lvtln_model.sh
sifeats="ark,s,cs:add-deltas scp:$sdata/JOB/feats.scp ark:- | apply-cmvn-sliding $cmvn_sliding_opts ark:- ark:- | select-voiced-frames ark:- scp,s,cs:$sdata/JOB/vad.scp ark:- | subsample-feats --n=$subsample ark:- ark:- |"
if [ $stage -le -1 ]; then
echo "$0: computing Gaussian selection info."
$cmd JOB=1:$nj $dir/log/gselect.JOB.log \
gmm-gselect --n=$num_gselect $vtlndir/final.ali_dubm "$sifeats" \
"ark:|gzip -c >$dir/gselect.JOB.gz" || exit 1;
fi
feats="$sifeats"
x=0
while [ $x -lt $num_iters ]; do
if [ $stage -le $x ]; then
echo "$0: pass $x of computing LVTLN transforms"
if [ $x -eq 0 ]; then ubm=$vtlndir/final.ali_dubm; else ubm=$vtlndir/final.dubm; fi
$cmd JOB=1:$nj $dir/log/lvtln.$x.JOB.log \
gmm-global-gselect-to-post "$ubm" "$feats" \
"ark,s,cs:gunzip -c $dir/gselect.JOB.gz|" ark:- \| \
gmm-global-est-lvtln-trans \
--logdet-scale=$logdet_scale --verbose=1 \
$vtlndir/final.dubm $vtlndir/final.lvtln "$sifeats" ark,s,cs:- \
ark:$dir/trans.$x.JOB ark,t:$dir/warp.$x.JOB || exit 1
# consolidate the warps into one file.
for j in $(seq $nj); do cat $dir/warp.$x.$j; done > $dir/warp.$x
rm $dir/warp.$x.*
fi
feats="$sifeats transform-feats ark:$dir/trans.$x.JOB ark:- ark:- |"
x=$[$x+1]
done
ln -sf warp.$[$x-1] $dir/utt2warp
$cleanup && rm $dir/gselect.*.gz $dir/trans.*
echo "$0: Distribution of classes for one job is below."
grep 'Distribution of classes' $dir/log/lvtln.$[$x-1].1.log
echo "$0: created warp factors in $dir/utt2warp"
# Summarize warning messages...
utils/summarize_warnings.pl $dir/log
echo "$0: Done getting VTLN warps in $dir"