(trunk/egs/gale_arabic) Resolving the sampling frequency mismatch issues

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4558 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

egs/gale_arabic/s5/local/gale_data_prep_audio.sh

@@ -46,13 +46,13 @@ done
 for w in `find $wavedir -name *.wav` ; do 
   base=`basename $w .wav`
   fullpath=`readlink -f $w`
-  echo "$base sox $fullpath -r 8000 -t wav - |"
+  echo "$base sox $fullpath -r 16000 -t wav - |"
 done
 
 for w in `find $wavedir -name *.flac` ; do 
   base=`basename $w .flac`
   fullpath=`readlink -f $w`
-  echo "$base sox $fullpath -r 8000 -t wav - |"
+  echo "$base sox $fullpath -r 16000 -t wav - |"
 done
 )  | sort -u > $galeData/wav.scp
 

egs/gale_arabic/s5/run.sh

@@ -82,7 +82,8 @@ local/gale_format_data.sh
 # want to store MFCC features.
 mfccdir=mfcc
 
-for x in test train; do
+for x in train test ; do
+  utils/fix_data_dir.sh data/$x # some files fail to get mfcc for many reasons
   steps/make_mfcc.sh --cmd "$train_cmd" --nj $nJobs \
     data/$x exp/make_mfcc/$x $mfccdir
   utils/fix_data_dir.sh data/$x # some files fail to get mfcc for many reasons