coderm

HTK应用程序

#!/bin/csh

# This script will encode a given list of files of the RM database
# store the results in a directory tree given the root
# It has options for coding training data and test data for both
# the speaker dependent and speaker independent training sets
#
# usage: coderm ndxdir cdmount dbtype dbportion targetroot config codelist
#
#  ndxdir       - directory to find the NIST filelists
#  cdmount      - where CD-ROM with both data files is mounted
#  dbtype       - ind | dep
#  dbportion    - train | dev_aug | traina | trainb | evalset
#  targetroot   - root directory for coded files
#  config       - configuration file used to set coding parameters
#  codelist     - list of coded files
#
#  for speaker dependent - dbportion = traina (1st 6 speakers)
#                                    = trainb (2nd 6 speakers)
#  for speaker independent - dbportion = train (main 72 speakers)
#                          - dbportion = dev_aug (augmented 37 speakers)
#  In both cases evalset is one of {feb89, oct89, feb91, sep92}
#
# eg coderm $RMLIB/ndx $CDRM1 ind train /home/azure/pcw/rmdata
#
# The script first identifies the appropriate filelist from ndxdir
# and then codes all the files listed transforming the filenames 
# so that they are unique e.g  bef0_3/st0013.wav -> bef0_3_st0013.mfc
#
# Copyright (c) Phil Woodland + Julian Odell, 1995
# Last Updated 17/8/95
#

if ( $#argv != 7 ) then
   echo "usage: coderm ndxdir cdmount dbtype dbportion targetroot config codelist"
   exit 1
endif

set ndxdir=$1
if ( ! -d $ndxdir ) then
   echo "NIST filelist directory $ndxdir not found"
   exit 2
endif

set rmcd=$2
if ( ! -d $rmcd ) then
   echo "RM CD-ROM $rmcd not found"
   exit 2
endif

set dbtype=$3
if ( $dbtype != ind && $dbtype != dep ) then
   echo "dbtype of [ind | dep] expected and not $dbtype"
   exit 2
endif

set dbportion=$4
switch ($dbportion)
case train: 
   if ($dbtype == ind) then
      set ndx=$ndxdir/72_indtr.ndx
   else
      echo "Illegal dbportion $dbportion for dbtype $dbtype"
      exit 2
   endif
   breaksw 
case dev_aug:
   if ($dbtype == ind) then
      set ndx=$ndxdir/37_indtr.ndx
   else
      echo "Illegal dbportion $dbportion for dbtype $dbtype"
      exit 2
   endif
   breaksw 
case traina: 
   if ($dbtype == dep) then
      set ndx=$ndxdir/6a_deptr.ndx
   else
      echo "Illegal dbportion $dbportion for dbtype $dbtype"
      exit 2
   endif
   breaksw 
case trainb: 
   if ($dbtype == dep) then
      set ndx=$ndxdir/6b_deptr.ndx
   else
      echo "Illegal dbportion $dbportion for dbtype $dbtype"
      exit 2
   endif
   breaksw 
case feb89:
   set ndx=$ndxdir/4_${dbtype}tst.ndx
   breaksw
case oct89:
   set ndx=$ndxdir/5_${dbtype}tst.ndx
   breaksw
case feb91:
   set ndx=$ndxdir/6_${dbtype}tst.ndx
   breaksw
case sep92:
   set ndx=$ndxdir/7_${dbtype}tst.ndx
   breaksw
case dms0_tst:
   set ndx=$ndxdir/dms0_tst.ndx
   breaksw
default: 
   echo "dbtype not one of (train | dev_aug | traina | trainb | feb89 | oct89 | feb91 | sep92 | dms0_tst)"
   exit 2
endsw

set tgtroot=$5
if ( ! -d $tgtroot ) then
   echo "Target root directory $tgtroot doesnt exist"
   exit 2
endif

set config=$6
set codefilelist=$7

if ( ! -r $ndx ) then
    echo "Cant read RM filelist $ndx"
    exit 2
endif

set codescript="${codefilelist}.code"

\rm -f $codefilelist
\rm -f $codescript

echo "Making file lists"
grep -v "^;" $ndx | sed -e "s:^/\(.*\)/\(.*\)\([/_]\)\(.*\)/\(.*\)/\(.*\)\.wav:${rmcd}/\1/\2\3\4/\5/\6.wav ${tgtroot}/\2_\4/\5/\5_\6.mfc:" > $codescript
awk '{ print $2}' $codescript > $codefilelist
set count=`cat $codefilelist | wc -l`

echo "Making directories"
foreach i (`sed 's:\(.*\)/[a-z0-9_]*\.mfc:\1:' $codefilelist | sort -u`)
   if ( ! -d $i ) then
      mkdir -p $i
      if ( $status != 0 ) then
         echo "Cannot Create Directory $i"
         exit 1
      endif
   endif
end

echo "Coding data"
HCopy -A -T 1 -C $config -S $codescript
if ( $status != 0 ) then
   echo "Bad Status after Coding"
   exit 1
endif
\rm -f $codescript
echo ""
echo "Coding complete - $count files processed"