Szu-JuiChen · Szu-JuiChen · Nov 7, 2018 · Nov 7, 2018 · Nov 7, 2018
diff --git a/egs/reverb/s5/RESULTS b/egs/reverb/s5/RESULTS
diff --git a/egs/reverb/s5/local/compute_se_scores.sh b/egs/reverb/s5/local/compute_se_scores.sh
@@ -0,0 +1,39 @@
+#!/bin/bash
+# Copyright 2018 Johns Hopkins University (Author: Aswin Shanmugam Subramanian)
+# Apache 2.0
+
+# This script computes the dereverberation scores given in REVERB challenge
+# Eg. local/compute_se_scores.sh --nch 8 /export/corpora5/REVERB_2014/REVERB ${PWD}/wav ${PWD}/local 
+
+. ./cmd.sh
+. ./path.sh
+set -e
+set -u
+set -o pipefail
+
+cmd=run.pl
+nch=8
+
+. utils/parse_options.sh || exit 1;
+
+if [ $# != 3 ]; then
+   echo "Wrong #arguments ($#, expected 3)"
+   echo "Usage: local/compute_se.sh [options] <reverb_data> <enhancement-directory> <pesq-directory>"
+   echo "options"
+   echo "  --cmd <cmd>                              # Command to run in parallel with"
+   echo "  --nch <nch>                              # nch of WPE to use for computing SE scores"
+   exit 1;
+fi
+
+reverb_data=$1
+enhancement_directory=$2
+pesqdir=$3
+enhancement_directory_sim=$enhancement_directory/WPE/${nch}ch/REVERB_WSJCAM0_dt/data/
+enhancement_directory_real=$enhancement_directory/WPE/${nch}ch/MC_WSJ_AV_Dev/
+
+expdir=${PWD}/exp/compute_se_${nch}ch
+pushd local/REVERB_scores_source/REVERB-SPEENHA.Release04Oct/evaltools
+$cmd $expdir/compute_se_real.log matlab -nodisplay -nosplash -r "addpath('SRMRToolbox'); score_RealData('$reverb_data','$enhancement_directory_real');exit"
+$cmd $expdir/compute_se_sim.log matlab -nodisplay -nosplash -r "addpath('SRMRToolbox'); score_SimData('$reverb_data','$enhancement_directory_sim','$pesqdir');exit"
+popd
+mv local/REVERB_scores_source/REVERB-SPEENHA.Release04Oct/scores $expdir/
diff --git a/egs/reverb/s5/local/download_se_eval_tool.sh b/egs/reverb/s5/local/download_se_eval_tool.sh
@@ -0,0 +1,33 @@
+#!/bin/bash
+# Copyright 2018 Johns Hopkins University (Author: Aswin Shanmugam Subramanian)
+# This script downloads the official REVERB challenge SE scripts and SRMR toolbox
+# This script also downloads and compiles PESQ
+# please make sure that you or your institution have the license to report PESQ
+# Apache 2.0
+
+wget 'https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-P.862-200102-I!!SOFT-ZST-E&type=items' -O PESQ.zip
+unzip PESQ.zip -d local/PESQ_sources
+rm PESQ.zip
+cd local/PESQ_sources/P862/Software/source
+gcc  *.c -lm -o PESQ
+cd ../../../../../
+mv local/PESQ_sources/P862/Software/source/PESQ local/
+
+wget 'https://reverb2014.dereverberation.com/tools/REVERB-SPEENHA.Release04Oct.zip' -O REVERB_scores.zip
+unzip REVERB_scores.zip -d local/REVERB_scores_source
+rm REVERB_scores.zip
+
+pushd local/REVERB_scores_source/REVERB-SPEENHA.Release04Oct/evaltools
+sed -i 's/wavread/audioread/g' prog/score_sim.m
+git clone https://github.com/MuSAELab/SRMRToolbox.git
+sed -i 's/wavread/audioread/g' SRMRToolbox/libs/preprocess.m
+sed -i 's/SRMR_main/SRMR/g' prog/score_real.m
+sed -i 's/SRMR_main/SRMR/g' prog/score_sim.m
+sed -i 's/+wb\ //g' prog/calcpesq.m
+sed -i 's/pesq_/_pesq_/g' prog/calcpesq.m
+sed -ie '30d;31d' prog/calcpesq.m
+patch score_RealData.m -i ../../../score_RealData.patch -o score_RealData_new.m
+mv score_RealData_new.m score_RealData.m
+patch score_SimData.m -i ../../../score_SimData.patch -o score_SimData_new.m
+mv score_SimData_new.m score_SimData.m
+popd
diff --git a/egs/reverb/s5/local/get_results.sh b/egs/reverb/s5/local/get_results.sh
@@ -15,6 +15,18 @@ echo ""
 echo "exp/tri3/decode_et_simu_1ch"
 cat exp/tri3/decode_et_simu_1ch/scoring_kaldi/best_wer*
 echo ""
+echo "exp/tri3/decode_dt_real_1ch_wpe"
+cat exp/tri3/decode_dt_real_1ch_wpe/scoring_kaldi/best_wer*
+echo ""
+echo "exp/tri3/decode_dt_simu_1ch_wpe"
+cat exp/tri3/decode_dt_simu_1ch_wpe/scoring_kaldi/best_wer*
+echo ""
+echo "exp/tri3/decode_et_real_1ch_wpe"
+cat exp/tri3/decode_et_real_1ch_wpe/scoring_kaldi/best_wer*
+echo ""
+echo "exp/tri3/decode_et_simu_1ch_wpe"
+cat exp/tri3/decode_et_simu_1ch_wpe/scoring_kaldi/best_wer*
+echo ""
 echo "exp/tri3/decode_dt_real_2ch_wpe"
 cat exp/tri3/decode_dt_real_2ch_wpe/scoring_kaldi/best_wer*
 echo ""
@@ -60,8 +72,11 @@ echo ""
 echo "exp/tri3/decode_et_real_8ch_beamformit"
 cat exp/tri3/decode_et_real_8ch_beamformit/scoring_kaldi/best_wer*
 echo ""
-echo "exp/tri3/decode_et_simu_8ch_beamformit"
-cat exp/tri3/decode_et_simu_8ch_beamformit/scoring_kaldi/best_wer*
+echo "exp/tri3/decode_dt_cln"
+cat exp/tri3/decode_dt_cln/scoring_kaldi/best_wer*
+echo ""
+echo "exp/tri3/decode_et_cln"
+cat exp/tri3/decode_et_cln/scoring_kaldi/best_wer*
 echo "########################################"
 echo "TDNN RESULTs:"
 echo "exp/chain_tr_simu_8ch/tdnn1a_sp/decode_test_tg_5k_dt*"

diff --git a/egs/reverb/s5/local/prepare_simu_data.sh b/egs/reverb/s5/local/prepare_simu_data.sh
@@ -54,6 +54,12 @@ for nch in 1 2 8; do
 	    perl -se 'while (<>) { chomp; if (m/\/(\w{8})[^\/]+$/) { print $1, " ", $dir, $_, "\n"; } }' -- -dir=${reverb}/REVERB_WSJCAM0_${task}/data ${taskdir}/$x |\
 		sed -e "s/^\(...\)/\1_${x}_\1/"
 	done > ${dir}/${task}_simu_${nch}ch_wav.scp
+	if [ ${nch} == 1 ]; then
+	    for x in `ls ${taskdir} | grep SimData | grep _${task}_ | grep -e cln`; do
+	        perl -se 'while (<>) { chomp; if (m/\/(\w{8})[^\/]+$/) { print $1, " ", $dir, $_, "\n"; } }' -- -dir=${reverb}/REVERB_WSJCAM0_${task}/data ${taskdir}/$x |\
+	    	sed -e "s/^\(...\)/\1_${x}_\1/"
+	    done > ${dir}/${task}_cln_wav.scp
+        fi
     done
 
     task=tr
@@ -83,13 +89,25 @@ for nch in 1 2 8; do
 		sed -e "s/^\(...\)/\1_${x}_\1/"
 	done > ${dir}/${task}_simu_${nch}ch.trans1 || exit 1;
 	cat ${dir}/${task}_simu_${nch}ch.trans1 | local/normalize_transcript.pl ${noiseword} > ${dir}/${task}_simu_${nch}ch.txt || exit 1;
+	if [ ${nch} == 1 ]; then
+	    for x in `ls ${taskdir} | grep SimData | grep _${task}_ | grep -e cln`; do
+		perl -e 'while (<>) { chomp; if (m/\/(\w{8})[^\/]+$/) { print $1, "\n"; } }' ${taskdir}/$x |\
+		    perl local/find_transcripts_singledot.pl ${dir}/${task}.dot |\
+		    sed -e "s/^\(...\)/\1_${x}_\1/"
+	    done > ${dir}/${task}_cln.trans1 || exit 1;
+	    cat ${dir}/${task}_cln.trans1 | local/normalize_transcript.pl ${noiseword} > ${dir}/${task}_cln.txt || exit 1;
+        fi
     done
 
     # Make the utt2spk and spk2utt files.
     for task in tr dt et; do
 	cat ${dir}/${task}_simu_${nch}ch_wav.scp | awk '{print $1}' | awk -F '_' '{print $0 " " $1}' > ${dir}/${task}_simu_${nch}ch.utt2spk || exit 1;
 	cat ${dir}/${task}_simu_${nch}ch.utt2spk | ./utils/utt2spk_to_spk2utt.pl > ${dir}/${task}_simu_${nch}ch.spk2utt || exit 1;
     done
+    for task in dt et; do
+	cat ${dir}/${task}_cln_wav.scp | awk '{print $1}' | awk -F '_' '{print $0 " " $1}' > ${dir}/${task}_cln.utt2spk || exit 1;
+	cat ${dir}/${task}_cln.utt2spk | ./utils/utt2spk_to_spk2utt.pl > ${dir}/${task}_cln.spk2utt || exit 1;
+    done
 done
 
 # finally copy the above files to the data directory
@@ -118,6 +136,14 @@ for nch in 1 2 8; do
 		sort ${dir}/${task}_simu_1ch.utt2spk > ${datadir}/utt2spk
 		sort ${dir}/${task}_simu_1ch.spk2utt > ${datadir}/spk2utt
 		./utils/fix_data_dir.sh ${datadir}
+            else
+		datadir=data/${task}_cln
+		mkdir -p ${datadir}
+		sort ${dir}/${task}_cln_wav.scp > ${datadir}/wav.scp
+		sort ${dir}/${task}_cln.txt     > ${datadir}/text
+		sort ${dir}/${task}_cln.utt2spk > ${datadir}/utt2spk
+		sort ${dir}/${task}_cln.spk2utt > ${datadir}/spk2utt
+		./utils/fix_data_dir.sh ${datadir}
 	    fi
 	fi
     done

diff --git a/egs/reverb/s5/local/score.sh b/egs/reverb/s5/local/score.sh
@@ -63,6 +63,9 @@ mkdir -p $dir/scoring_kaldi
 if echo $data | grep -q "real"; then
   tasks="\
   near_room1 far_room1"
+elif echo $data | grep -q "cln"; then
+  tasks="\
+  cln_room1 cln_room2 cln_room3"
 else
   tasks="\
   near_room1 far_room1 \

diff --git a/egs/reverb/s5/run.sh b/egs/reverb/s5/run.sh
@@ -36,6 +36,10 @@ fi
 . ./path.sh
 
 stage=0
+nch_se=8
+# flag for turing on computation of dereverberation measures
+# please make sure that you or your institution have the license to report PESQ before turning on the flag
+compute_se=false
 
 . utils/parse_options.sh
 # Set bash to 'debug' mode, it prints the commands (option '-x') and exits on :
@@ -57,7 +61,7 @@ fi
 
 #training set and test set
 train_set=tr_simu_8ch
-test_sets="dt_real_1ch dt_simu_1ch et_real_1ch et_simu_1ch dt_real_2ch_beamformit dt_simu_2ch_beamformit et_real_2ch_beamformit et_simu_2ch_beamformit dt_real_8ch_beamformit dt_simu_8ch_beamformit et_real_8ch_beamformit et_simu_8ch_beamformit dt_real_1ch_wpe dt_simu_1ch_wpe et_real_1ch_wpe et_simu_1ch_wpe dt_real_2ch_wpe dt_simu_2ch_wpe et_real_2ch_wpe et_simu_2ch_wpe dt_real_8ch_wpe dt_simu_8ch_wpe et_real_8ch_wpe et_simu_8ch_wpe"
+test_sets="dt_real_1ch dt_simu_1ch et_real_1ch et_simu_1ch dt_real_2ch_beamformit dt_simu_2ch_beamformit et_real_2ch_beamformit et_simu_2ch_beamformit dt_real_8ch_beamformit dt_simu_8ch_beamformit et_real_8ch_beamformit et_simu_8ch_beamformit dt_real_1ch_wpe dt_simu_1ch_wpe et_real_1ch_wpe et_simu_1ch_wpe dt_real_2ch_wpe dt_simu_2ch_wpe et_real_2ch_wpe et_simu_2ch_wpe dt_real_8ch_wpe dt_simu_8ch_wpe et_real_8ch_wpe et_simu_8ch_wpe dt_cln et_cln"
 
 # The language models with which to decode (tg_5k or bg_5k)
 lm="tg_5k"
@@ -68,6 +72,7 @@ nj=92
 decode_nj=10
 
 wavdir=${PWD}/wav
+pesqdir=${PWD}/local
 if [ ${stage} -le 1 ]; then
   # data preparation
   echo "stage 0: Data preparation"
@@ -81,7 +86,18 @@ if [ $stage -le 2 ]; then
   local/run_beamform.sh ${wavdir}/WPE/
 fi
 
-if [ $stage -le 3 ]; then
+# Compute dereverberation scores
+if [ $stage -le 3 ] && $compute_se; then
+  if [ ! -d local/REVERB_scores_source ] || [ ! -d local/REVERB_scores_source/REVERB-SPEENHA.Release04Oct/evaltools/SRMRToolbox ] || [ ! -f local/PESQ ]; then
+    # download and install speech enhancement evaluation tools
+    local/download_se_eval_tool.sh
+  fi
+  local/compute_se_scores.sh --nch $nch_se $reverb $wavdir $pesqdir
+  cat exp/compute_se_${nch_se}ch/scores/score_SimData
+  cat exp/compute_se_${nch_se}ch/scores/score_RealData
+fi
+
+if [ $stage -le 4 ]; then
   # Prepare wsjcam0 clean data and wsj0 language model.
   local/wsjcam0_data_prep.sh $wsjcam0 $wsj0
 
@@ -105,14 +121,14 @@ if [ $stage -le 3 ]; then
 		data/lang $LM data/local/dict/lexicon.txt data/lang
 fi
 
-if [ $stage -le 4 ]; then
+if [ $stage -le 5 ]; then
   for dset in ${train_set} ${test_sets}; do
     utils/copy_data_dir.sh data/${dset} data/${dset}_nosplit
     utils/data/modify_speaker_info.sh --seconds-per-spk-max 180 data/${dset}_nosplit data/${dset}
   done
 fi
 
-if [ $stage -le 5 ]; then
+if [ $stage -le 6 ]; then
   # Extract MFCC features for train and test sets.
   mfccdir=mfcc
   for x in ${train_set} ${test_sets}; do
@@ -122,29 +138,29 @@ if [ $stage -le 5 ]; then
   done
 fi
 
-if [ $stage -le 6 ]; then
+if [ $stage -le 7 ]; then
   # Starting basic training on MFCC features
   steps/train_mono.sh --nj $nj --cmd "$train_cmd" \
 		      data/${train_set} data/lang exp/mono
 fi
 
-if [ $stage -le 7 ]; then
+if [ $stage -le 8 ]; then
   steps/align_si.sh --nj $nj --cmd "$train_cmd" \
 		    data/${train_set} data/lang exp/mono exp/mono_ali
 
   steps/train_deltas.sh --cmd "$train_cmd" \
 			2500 30000 data/${train_set} data/lang exp/mono_ali exp/tri1
 fi
 
-if [ $stage -le 8 ]; then
+if [ $stage -le 9 ]; then
   steps/align_si.sh --nj $nj --cmd "$train_cmd" \
 		    data/${train_set} data/lang exp/tri1 exp/tri1_ali
 
   steps/train_lda_mllt.sh --cmd "$train_cmd" \
 			  4000 50000 data/${train_set} data/lang exp/tri1_ali exp/tri2
 fi
 
-if [ $stage -le 9 ]; then
+if [ $stage -le 10 ]; then
   utils/mkgraph.sh data/lang_test_$lm exp/tri2 exp/tri2/graph
   for dset in ${test_sets}; do
     steps/decode.sh --nj $decode_nj --cmd "$decode_cmd"  --num-threads 4 \
@@ -153,15 +169,15 @@ if [ $stage -le 9 ]; then
   wait
 fi
 
-if [ $stage -le 10 ]; then
+if [ $stage -le 11 ]; then
   steps/align_si.sh --nj $nj --cmd "$train_cmd" \
 		    data/${train_set} data/lang exp/tri2 exp/tri2_ali
 
   steps/train_sat.sh --cmd "$train_cmd" \
 		     5000 100000 data/${train_set} data/lang exp/tri2_ali exp/tri3
 fi
 
-if [ $stage -le 11 ]; then
+if [ $stage -le 12 ]; then
   utils/mkgraph.sh data/lang_test_$lm exp/tri3 exp/tri3/graph
   for dset in ${test_sets}; do
     steps/decode_fmllr.sh --nj $decode_nj --cmd "$decode_cmd"  --num-threads 4 \
@@ -170,13 +186,13 @@ if [ $stage -le 11 ]; then
   wait
 fi
 
-if [ $stage -le 12 ]; then
+if [ $stage -le 13 ]; then
   # chain TDNN
   local/chain/run_tdnn.sh --nj ${nj} --train-set ${train_set} --test-sets "$test_sets" --gmm tri3 --nnet3-affix _${train_set} \
   --lm-suffix _test_$lm
 fi
 
 # get all WERs. 
-if [ $stage -le 13 ]; then
+if [ $stage -le 14 ]; then
   local/get_results.sh
 fi