bool CSoundCombiner::LoadSentenceFromWavFileUsingIO( char const *wavfile, CSentence& sentence, IFileReadBinary& io )
{
sentence.Reset();
InFileRIFF riff( wavfile, io );
// UNDONE: Don't use printf to handle errors
if ( riff.RIFFName() != RIFF_WAVE )
{
return false;
}
// set up the iterator for the whole file (root RIFF is a chunk)
IterateRIFF walk( riff, riff.RIFFSize() );
// This chunk must be first as it contains the wave's format
// break out when we've parsed it
bool found = false;
while ( walk.ChunkAvailable() && !found )
{
switch( walk.ChunkName() )
{
case WAVE_VALVEDATA:
{
found = true;
CSoundCombiner::ParseSentence( sentence, walk );
}
break;
}
walk.ChunkNext();
}
return true;
}
//-----------------------------------------------------------------------------
// Purpose: Given a wavfile and a list of inwords, determines the word/phonene
// sample counts for the sentce
// Input : *wavfile -
// *inwords -
// *outphonemes{ text.Clear( -
// Output : SR_RESULT
//-----------------------------------------------------------------------------
static SR_RESULT SAPI_ExtractPhonemes(
const char *wavfile,
int numsamples,
void (*pfnPrint)( const char *fmt, ... ),
CSentence& inwords,
CSentence& outwords )
{
LogReset();
USES_CONVERSION;
CSpDynamicString text;
text.Clear();
HKEY hkwipe;
LONG lResult = RegOpenKeyEx( HKEY_CURRENT_USER, "Software\\Microsoft\\Speech\\RecoProfiles", 0, KEY_ALL_ACCESS, &hkwipe );
if ( lResult == ERROR_SUCCESS )
{
RecursiveRegDelKey( hkwipe );
RegCloseKey( hkwipe );
}
if ( strlen( inwords.GetText() ) <= 0 )
{
inwords.SetTextFromWords();
}
// Construct a string from the inwords array
text.Append( T2W( inwords.GetText() ) );
// Assume failure
SR_RESULT result = SR_RESULT_ERROR;
if ( text.Length() > 0 )
{
CSentence sentence;
pfnPrint( "Processing...\r\n" );
// Give it a try
result = ExtractPhonemes( wavfile, text, sentence, pfnPrint );
pfnPrint( "Finished.\r\n" );
// PrintWordsAndPhonemes( sentence, pfnPrint );
// Copy results to outputs
outwords.Reset();
outwords.SetText( inwords.GetText() );
Log( "Starting\n" );
LogWords( inwords );
if ( SR_RESULT_ERROR != result )
{
int i;
Log( "Hypothesized\n" );
LogWords( sentence );
for( i = 0 ; i < sentence.m_Words.Size(); i++ )
{
CWordTag *tag = sentence.m_Words[ i ];
if ( tag )
{
// Skip '...' tag
if ( stricmp( tag->GetWord(), "..." ) )
{
CWordTag *newTag = new CWordTag( *tag );
outwords.m_Words.AddToTail( newTag );
}
}
}
// Now insert unrecognized/skipped words from original list
//
int frompos = 0, topos = 0;
while( 1 )
{
// End of source list
if ( frompos >= inwords.m_Words.Size() )
break;
const CWordTag *fromTag = inwords.m_Words[ frompos ];
// Reached end of destination list, just copy words over from from source list until
// we run out of source words
if ( topos >= outwords.m_Words.Size() )
{
// Just copy words over
CWordTag *newWord = new CWordTag( *fromTag );
// Remove phonemes
while ( newWord->m_Phonemes.Size() > 0 )
{
CPhonemeTag *kill = newWord->m_Phonemes[ 0 ];
newWord->m_Phonemes.Remove( 0 );
delete kill;
}
outwords.m_Words.AddToTail( newWord );
frompos++;
topos++;
continue;
}
// Destination word
const CWordTag *toTag = outwords.m_Words[ topos ];
// Words match, just skip ahead
if ( !stricmp( fromTag->GetWord(), toTag->GetWord() ) )
{
frompos++;
topos++;
continue;
}
// The only case we handle is that something in the source wasn't in the destination
// Find the next source word that appears in the destination
int skipAhead = frompos + 1;
bool found = false;
while ( skipAhead < inwords.m_Words.Size() )
{
const CWordTag *sourceWord = inwords.m_Words[ skipAhead ];
if ( !stricmp( sourceWord->GetWord(), toTag->GetWord() ) )
{
found = true;
break;
}
skipAhead++;
}
// Uh oh destination has words that are not in source, just skip to next destination word?
if ( !found )
{
topos++;
}
else
{
// Copy words from from source list into destination
//
int skipCount = skipAhead - frompos;
while ( --skipCount>= 0 )
{
const CWordTag *sourceWord = inwords.m_Words[ frompos++ ];
CWordTag *newWord = new CWordTag( *sourceWord );
// Remove phonemes
while ( newWord->m_Phonemes.Size() > 0 )
{
CPhonemeTag *kill = newWord->m_Phonemes[ 0 ];
newWord->m_Phonemes.Remove( 0 );
delete kill;
}
outwords.m_Words.InsertBefore( topos, newWord );
topos++;
}
frompos++;
topos++;
}
}
Log( "\nDone simple check\n" );
LogWords( outwords );
LogPhonemes( outwords );
ComputeMissingByteSpans( numsamples, outwords );
Log( "\nFinal check\n" );
LogWords( outwords );
LogPhonemes( outwords );
}
}
else
{
pfnPrint( "Input sentence is empty!\n" );
}
// Return results
return result;
}