unsigned int CSentence::ComputeDataCheckSum()
{
#if PHONEME_EDITOR
int i;
int c;
CRC32_t crc;
CRC32_Init( &crc );
// Checksum the text
CRC32_ProcessBuffer( &crc, GetText(), Q_strlen( GetText() ) );
// Checsum words and phonemes
c = m_Words.Count();
for ( i = 0; i < c; ++i )
{
CWordTag *word = m_Words[ i ];
unsigned int wordCheckSum = word->ComputeDataCheckSum();
CRC32_ProcessBuffer( &crc, &wordCheckSum, sizeof( unsigned int ) );
}
// Checksum emphasis data
c = m_EmphasisSamples.Count();
for ( i = 0; i < c; ++i )
{
CRC32_ProcessBuffer( &crc, &m_EmphasisSamples[ i ].time, sizeof( float ) );
CRC32_ProcessBuffer( &crc, &m_EmphasisSamples[ i ].value, sizeof( float ) );
}
CRC32_Final( &crc );
return ( unsigned int )crc;
#else
Assert( 0 );
return 0;
#endif
}
void PrintWordsAndPhonemes( CSentence& sentence, void (*pfnPrint)( const char *fmt, ... ) )
{
char sz[ 256 ];
int i;
pfnPrint( "WORDS\r\n\r\n" );
for ( i = 0 ; i < sentence.m_Words.Size(); i++ )
{
CWordTag *word = sentence.m_Words[ i ];
if ( !word )
continue;
sprintf( sz, "<%u - %u> %s\r\n",
word->m_uiStartByte, word->m_uiEndByte, word->GetWord() );
pfnPrint( sz );
for ( int j = 0 ; j < word->m_Phonemes.Size(); j++ )
{
CPhonemeTag *phoneme = word->m_Phonemes[ j ];
if ( !phoneme )
continue;
sprintf( sz, " <%u - %u> %s\r\n",
phoneme->m_uiStartByte, phoneme->m_uiEndByte, phoneme->GetTag() );
pfnPrint( sz );
}
}
pfnPrint( "\r\n" );
}
//-----------------------------------------------------------------------------
// Purpose:
// Input : from -
//-----------------------------------------------------------------------------
CWordTag::CWordTag( const CWordTag& from )
{
m_pszWord = NULL;
SetWord( from.m_pszWord );
SetStartAndEndBytes( from.GetStartByte(), from.GetEndByte() );
m_flStartTime = from.m_flStartTime;
m_flEndTime = from.m_flEndTime;
SetSelected( from.GetSelected() );
for ( int p = 0; p < from.m_Phonemes.Size(); p++ )
{
CPhonemeTag *newPhoneme = new CPhonemeTag( *from.m_Phonemes[ p ] );
m_Phonemes.AddToTail( newPhoneme );
}
}
//-----------------------------------------------------------------------------
// Purpose:
//-----------------------------------------------------------------------------
void CSentence::SetTextFromWords( void )
{
#if PHONEME_EDITOR
char fulltext[ 1024 ];
fulltext[ 0 ] = 0;
for ( int i = 0 ; i < m_Words.Size(); i++ )
{
CWordTag *word = m_Words[ i ];
Q_strncat( fulltext, word->GetWord(), sizeof( fulltext ), COPY_ALL_CHARACTERS );
if ( i != m_Words.Size() )
{
Q_strncat( fulltext, " ", sizeof( fulltext ), COPY_ALL_CHARACTERS );
}
}
SetText( fulltext );
#endif
}
void CSentence::SaveToBuffer( CUtlBuffer& buf )
{
#if PHONEME_EDITOR
Assert( !m_bIsCached );
int i, j;
buf.Printf( "VERSION 1.0\n" );
buf.Printf( "PLAINTEXT\n" );
buf.Printf( "{\n" );
buf.Printf( "%s\n", GetText() );
buf.Printf( "}\n" );
buf.Printf( "WORDS\n" );
buf.Printf( "{\n" );
for ( i = 0; i < m_Words.Size(); i++ )
{
CWordTag *word = m_Words[ i ];
Assert( word );
buf.Printf( "WORD %s %.3f %.3f\n",
word->GetWord(),
word->m_flStartTime,
word->m_flEndTime );
buf.Printf( "{\n" );
for ( j = 0; j < word->m_Phonemes.Size(); j++ )
{
CPhonemeTag *phoneme = word->m_Phonemes[ j ];
Assert( phoneme );
buf.Printf( "%i %s %.3f %.3f 1\n",
phoneme->GetPhonemeCode(),
phoneme->GetTag(),
phoneme->GetStartTime(),
phoneme->GetEndTime() );
}
buf.Printf( "}\n" );
}
buf.Printf( "}\n" );
buf.Printf( "EMPHASIS\n" );
buf.Printf( "{\n" );
int c = m_EmphasisSamples.Count();
for ( i = 0; i < c; i++ )
{
CEmphasisSample *sample = &m_EmphasisSamples[ i ];
Assert( sample );
buf.Printf( "%f %f\n", sample->time, sample->value );
}
buf.Printf( "}\n" );
buf.Printf( "OPTIONS\n" );
buf.Printf( "{\n" );
buf.Printf( "voice_duck %d\n", GetVoiceDuck() ? 1 : 0 );
if ( m_bStoreCheckSum )
{
buf.Printf( "checksum %d\n", m_uCheckSum );
}
buf.Printf( "}\n" );
#else
Assert( 0 );
#endif
}
void CSentence::CreateEventWordDistribution( char const *pszText, float flSentenceDuration )
{
Assert( pszText );
if ( !pszText )
return;
int wordCount = CountWords( pszText );
if ( wordCount <= 0 )
return;
float wordLength = ( flSentenceDuration - 2 * STARTEND_TIMEGAP) / (float)wordCount;
float wordStart = STARTEND_TIMEGAP;
Reset();
char word[ 256 ];
unsigned char const *in = (unsigned char *)pszText;
char *out = word;
while ( *in )
{
if ( !ShouldSplitWord( *in ) )
{
*out++ = *in++;
}
else
{
*out = 0;
// Skip over splitters
while ( *in && ( ShouldSplitWord( *in ) ) )
{
in++;
}
if ( strlen( word ) > 0 )
{
CWordTag *w = new CWordTag();
Assert( w );
w->SetWord( word );
w->m_flStartTime = wordStart;
w->m_flEndTime = wordStart + wordLength;
AddWordTag( w );
wordStart += wordLength;
}
out = word;
}
}
*out = 0;
if ( strlen( word ) > 0 )
{
CWordTag *w = new CWordTag();
Assert( w );
w->SetWord( word );
w->m_flStartTime = wordStart;
w->m_flEndTime = wordStart + wordLength;
AddWordTag( w );
wordStart += wordLength;
}
}
//-----------------------------------------------------------------------------
// 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;
}