kaldi中腳本東西比較多，一層嵌一層，不易閱讀。
本文以yesno為例，直接使用kaldi編譯的工具，書寫簡易訓練步驟，方便學習kaldi工具的使用。
注意：轉載請注明出處。
yesno訓練
- 准備數據
- 在yesno/s5下新建文件夾：
mkdir easy，后續的操作將在easy文件夾中執行。 - 拷貝s5下
./path到easy文件夾中，./path的作用是能直接調用工具，不用添加工具所在路徑，類似於設置環境變量。 - 本腳主要便於理解kaldi工具的使用，一些批處理和數據下載並沒有做，需要運行一遍
yesno/s5/./run.sh生成訓練所需輸入。 - 到
s5/data/train下拷貝wav.scp到easy目錄下作為訓練輸入，因為wav.scp是相對路徑也需要拷貝waves_yesno/到easy下。 - 拷貝詞典到目錄下：拷貝
s5/input到easy目錄下。
准備數據結束，可以寫自己的腳本了。
- 在yesno/s5下新建文件夾：
先給出整體腳本如下：
#!/bin/bash
. ./path
# feature extraction:
# a series of light command [ compute-mfcc + copy-feats -> compute-cmvn-stats -> apply-cmvn -> add-deltas ]
# the data flow transition [ wav -> mfcc.ark,scp -> cmvn.ark,scp -> delta.ark ]
mkdir mfcc
compute-mfcc-feats --verbose=2 --config="../conf/mfcc.conf" scp,p:wav.scp ark:- | copy-feats --compress=true ark:- ark,scp:mfcc/mfcc.ark,mfcc/mfcc.scp
compute-cmvn-stats scp:mfcc/mfcc.scp ark:mfcc/cmvn.ark
apply-cmvn ark:mfcc/cmvn.ark scp:mfcc/mfcc.scp ark:- | add-deltas ark:- ark:mfcc/delta.ark
# prepare dict for lang:
# input data [ lexicon_nosil.txt lexicon.txt phones.txt ]
# output data [ lexicon.txt lexicon_words.txt nonsilence_phones.txt optional_silence.txt silence_phones.txt ]
mkdir -p lang/dict
cp input/lexicon_nosil.txt lang/dict/lexicon_words.txt
cp input/lexicon.txt lang/dict/lexicon.txt
cat input/phones.txt | grep -v SIL > lang/dict/nonsilence_phones.txt
echo "SIL" > lang/dict/silence_phones.txt
echo "SIL" > lang/dict/optional_silence.txt
echo "Dictionary preparation succeeded"
# generate [ topo ] for acoustic model
utils/gen_topo.pl 3 5 2:3 1 > lang/lang/topo
# from [lexicoin phone word] -> [L.fst word.txt] for [G.fst train.fst HCLG.fst]
utils/prepare_lang.sh --position-dependent-phones false lang/dict "<SIL>" lang/local lang/lang
# train monophic acoustic model
# 1.from [topo 39] -> 0.mdl tree
gmm-init-mono --train-feats=ark:mfcc/delta.ark lang/lang/topo 39 mono/0.mdl mono/tree
# 2.from [L.fst 0.mdl tree word.txt text] -> train.fst
# compile-train-graphs [options] <tree-in> <model-in> <lexicon-fst-in> <transcriptions-rspecifier> <graphs-wspecifier>
compile-train-graphs mono/tree mono/0.mdl lang/lang/L.fst 'ark:sym2int.pl -f 2- lang/lang/words.txt text|' ark:lang/lang/graphs.fsts
# 3.from [graphs.fst] equally align the train data -> [ euqal.ali ]
# align-equal-compiled <graphs-rspecifier> <features-rspecifier> <alignments-wspecifier>
align-equal-compiled ark:lang/lang/graphs.fsts ark:mfcc/delta.ark ark:mono/equal.ali
# 4.from [equal.ali delta.ark mdl] -> [ 0.acc ]
gmm-acc-stats-ali mono/0.mdl ark:mfcc/delta.ark ark:mono/equal.ali mono/0.acc
# 5.from [0.mdl 0.acc] -> [ 1.mdl ]
# parameter est:
gmm-est mono/0.mdl mono/0.acc mono/1.mdl
x=1
numliter=40
numgauss=11
while [ $x -lt $numliter ]; do
# 6.from [1.mdl graphs.fst] align the data by new model -> [ 1.ali ]
gmm-align-compiled --beam=6 --retry-beam=20 mono/$x.mdl ark:lang/lang/graphs.fsts ark:mfcc/delta.ark ark:mono/$x.ali
# 4.from [equal.ali delta.ark mdl] -> [ 0.acc ]
gmm-acc-stats-ali mono/$x.mdl ark:mfcc/delta.ark ark:mono/equal.ali mono/$x.acc
# 5.from [x.mdl x.acc] -> [ x+1.mdl ]
gmm-est --mix-up=$numgauss --power=0.25 mono/$x.mdl mono/$x.acc mono/$[$x+1].mdl
numgauss=$[$numgauss+25]
x=$[$x+1]
done
cp mono/$x.mdl mono/final.mdl
# Graph compilation
# from [input/task.arpabo word.txt] -> G.fst
arpa2fst --disambig-symbol=#0 --read-symbol-table=lang/lang/words.txt input/task.arpabo lang/lang/G.fst
fstisstochastic lang/lang/G.fst
# from [final.mdl G.fst L.fst tree] -> HLCG.fst
utils/mkgraph.sh lang/lang mono mono/graph
分塊詳解
首先進行特征提取：
#!/bin/bash
. ./path
# 特征提取: compute-mfcc-feats, copy-feats
# 輸入為：wav.scp 輸出為:mfcc.ark,mfcc.scp
compute-mfcc-feats --verbose=2 --config="../conf/mfcc.conf" scp,p:wav.scp ark:- | copy-feats --compress=true ark:- ark,scp:mfcc/mfcc.ark,mfcc/mfcc.scp
# 計算均方歸一化矩陣：
# 輸入為：mfcc.ark,mfcc.scp 輸出為：mfcc/cmvn.ark,mfcc/cmvn.scp
compute-cmvn-stats scp:mfcc/mfcc.scp ark,scp:mfcc/cmvn.ark,mfcc/cmvn.scp
# 計算一階二階差分：
# 輸入為：mfcc/cmvn.ark,mfcc/cmvn.scp 輸出為：delta.ark
apply-cmvn scp:mfcc/cmvn.scp scp:mfcc/mfcc.scp ark:- | add-deltas ark:- ark:mfcc/delta.ark
然后，准備訓練所需的詞典，音素文件，詞文件等。
yesno里准備好了，直接拷貝即可。
# prepare dict for lang:
# input data [ lexicon_nosil.txt lexicon.txt phones.txt ]
# output data [ lexicon.txt lexicon_words.txt nonsilence_phones.txt optional_silence.txt silence_phones.txt ]
mkdir -p lang/dict
cp input/lexicon_nosil.txt lang/dict/lexicon_words.txt
cp input/lexicon.txt lang/dict/lexicon.txt
cat input/phones.txt | grep -v SIL > lang/dict/nonsilence_phones.txt
echo "SIL" > lang/dict/silence_phones.txt
echo "SIL" > lang/dict/optional_silence.txt
echo "Dictionary preparation succeeded"
生成聲學拓撲結構。
生成 L.fst word.txt用來生成G.fst train.fst HCLG.fst。其中utils/prepare_lang.sh所需全部輸入為上一步生成的dict文件。
# generate [ topo ] for acoustic model
utils/gen_topo.pl 3 5 2:3 1 > lang/lang/topo
# from [lexicoin phone word] -> [L.fst word.txt] for [G.fst train.fst HCLG.fst]
utils/prepare_lang.sh --position-dependent-phones false lang/dict "<SIL>" lang/local lang/lang
訓練單音素模型
- 流程如下 ：
- 利用生成的聲學拓撲初始化模型
- 生成訓練圖
- 初始化對齊
- 生成統計量
- 模型參數估計
- {重新對齊生，成統計量，模型參數估計}x10
- 生成並導出最終模型：
# train monophic acoustic model
# 1.from [topo 39] -> 0.mdl tree
gmm-init-mono --train-feats=ark:mfcc/delta.ark lang/lang/topo 39 mono/0.mdl mono/tree
# 2.from [L.fst 0.mdl tree word.txt text] -> train.fst
# compile-train-graphs [options] <tree-in> <model-in> <lexicon-fst-in> <transcriptions-rspecifier> <graphs-wspecifier>
compile-train-graphs mono/tree mono/0.mdl lang/lang/L.fst 'ark:sym2int.pl -f 2- lang/lang/words.txt text|' ark:lang/lang/graphs.fsts
# 3.from [graphs.fst] equally align the train data -> [ euqal.ali ]
# align-equal-compiled <graphs-rspecifier> <features-rspecifier> <alignments-wspecifier>
align-equal-compiled ark:lang/lang/graphs.fsts ark:mfcc/delta.ark ark:mono/equal.ali
# 4.from [equal.ali delta.ark mdl] -> [ 0.acc ]
gmm-acc-stats-ali mono/0.mdl ark:mfcc/delta.ark ark:mono/equal.ali mono/0.acc
# 5.from [0.mdl 0.acc] -> [ 1.mdl ]
# parameter est:
gmm-est mono/0.mdl mono/0.acc mono/1.mdl
x=1
numliter=40
numgauss=11
while [ $x -lt $numliter ]; do
# 6.from [1.mdl graphs.fst] align the data by new model -> [ 1.ali ]
gmm-align-compiled --beam=6 --retry-beam=20 mono/$x.mdl ark:lang/lang/graphs.fsts ark:mfcc/delta.ark ark:mono/$x.ali
# 4.from [equal.ali delta.ark mdl] -> [ 0.acc ]
gmm-acc-stats-ali mono/$x.mdl ark:mfcc/delta.ark ark:mono/equal.ali mono/$x.acc
# 5.from [x.mdl x.acc] -> [ x+1.mdl ]
gmm-est --mix-up=$numgauss --power=0.25 mono/$x.mdl mono/$x.acc mono/$[$x+1].mdl
numgauss=$[$numgauss+25]
x=$[$x+1]
done
cp mono/$x.mdl mono/final.mdl
最后合成語言模型：
# Graph compilation
# from [input/task.arpabo word.txt] -> G.fst
arpa2fst --disambig-symbol=#0 --read-symbol-table=lang/lang/words.txt input/task.arpabo lang/lang/G.fst
fstisstochastic lang/lang/G.fst
# from [final.mdl G.fst L.fst tree] -> HLCG.fst
utils/mkgraph.sh lang/lang mono mono/graph
運行結果：
建立解碼腳本
解碼指令較簡單一個指令即可：
#Usage: gmm-latgen-faster [options] model-in (fst-in|fsts-rspecifier) features-rspecifier lattice-wspecifier [ words-wspecifier [alignments-wspecifier] ]
gmm-latgen-faster --max-active=7000 --beam=13 --lattice-beam=6 --acoustic-scale=0.083333 \
--allow-partial=true --word-symbol-table=lang/lang/words.txt mono/final.mdl \
mono/graph/HCLG.fst ark:mfcc/delta.ark "ark:|gzip -c > result/lat.gz"
可以得到識別結果不是很好，沒關系，主要用這個例子來理解kaldi是怎么樣使用工具的。
轉載請注明出處：https://blog.csdn.net/chinatelecom08/article/details/81392399