audiosource.cpp

VC++视频开发实例集锦(包括“远程视频监控”"语音识别系统"等13个经典例子)

			,BOS=2
			,EOS=4
		};

		char Magic[4]; // The String "OggS" in ASCII
		uint8 Version; // Currently 0
		uint8 Flags; // Enf of stream, Beginning of stream
		uint64 GranulePos; // Not used for header packets. This is big endian !!!
		uint32 Serial; // serial number of the stream used for multiplexing/chaining
		uint32 PageNumber; // Page number within the stream
		uint32 CRC; // CRC value, similar to the ZIP CRC32, but with swapped bits. For more info I can provide a small Java class.

		uint8 LacingCount;
	 };
#pragma pack()

	 uint8*HeaderStart[3]={Header[0],Header[1],Header[2]};
	 const uint8*BufferStart=Buffer;

	 for(size_t PacketIndex=0;PacketIndex<3;)
	 {
		 // Read the next header
		 if(BufferSize<sizeof(PageHeader))
			 return false; // No data available for fixed header

		 // Read the fixed header
		 const PageHeader*CurrentPageHeader=reinterpret_cast<const PageHeader*>(Buffer);
		 Buffer+=sizeof(PageHeader);
		 BufferSize-=sizeof(PageHeader);

		 // Verify some fields
		 if(CurrentPageHeader->Magic[0]!='O'
			 ||CurrentPageHeader->Magic[1]!='g'
			 ||CurrentPageHeader->Magic[2]!='g'
			 ||CurrentPageHeader->Magic[3]!='S'
			 )
			 return false; // Our minimum requirement failed, see above for details

		 if(CurrentPageHeader->LacingCount>BufferSize)
			 return false; // No data available for lacing values
		 const uint8*CurrentLacings=Buffer;
		 Buffer+=CurrentPageHeader->LacingCount;
		 BufferSize-=CurrentPageHeader->LacingCount;

		 for(uint8 LacingIndex=0;PacketIndex<3&&LacingIndex<CurrentPageHeader->LacingCount;)
		 {
			 for(;PacketIndex<3&&LacingIndex<CurrentPageHeader->LacingCount;++LacingIndex)
			 {
				 if(BufferSize<CurrentLacings[LacingIndex])
					 return false; // No more data for the current part

				 // Copy the data
				 if(HeaderSize[PacketIndex]<CurrentLacings[LacingIndex])
					 return false; // No more space in output buffer
				 memcpy(Header[PacketIndex],Buffer,CurrentLacings[LacingIndex]);
				 HeaderSize[PacketIndex]-=CurrentLacings[LacingIndex];
				 Header[PacketIndex]+=CurrentLacings[LacingIndex];
				 Buffer+=CurrentLacings[LacingIndex];

				 if(CurrentLacings[LacingIndex]<255)
				 {
					 // Found new end of packet. If this is the last header
					 // packet, we must be at the page end.
					 if(PacketIndex==2&&LacingIndex<(CurrentPageHeader->LacingCount-1))
						 return false;

					 ++PacketIndex;
				 }
			 }
		 }
	 }

	 // Buffer points to the end of the last header page
	 SplitPosition=Buffer-BufferStart;
	 for(size_t i=0;i<3;++i)
	 {
		HeaderSize[i]=Header[i]-HeaderStart[i];
		Header[i]=HeaderStart[i];
	 }

	 return true;
}


BOOL AudioSourceOggVorbis::init() {

	if( !f ) return FALSE;

	char	*buffer;
	int		bytes;

	ogg_sync_init(&oy); /* Now we can read pages */
	
	if( streamInit()==FALSE ) return FALSE;

	fseek( f, 0, SEEK_SET );
	uint8* Buffer = (uint8*)malloc( pos );
	size_t Size = fread( Buffer, 1, pos, f );

	uint8* Header = (uint8*)malloc( pos );
	uint8* Comments = (uint8*)malloc( pos );
	uint8* Codebook = (uint8*)malloc( pos );

	if( Buffer==NULL || Header==NULL || Comments==NULL || Codebook==NULL ) {
failed:
		if( Buffer ) free( Buffer );
		if( Header ) free( Header );
		if( Comments ) free( Comments );
		if( Codebook ) free( Codebook );
		return FALSE;
	}

	size_t SplitPosition;
	size_t PacketSizes[] = { pos, pos, pos };
	uint8* PacketAddresses[] = { Header, Comments, Codebook };

	if( splitHeader( Buffer, Size, SplitPosition, PacketAddresses, PacketSizes )==FALSE )
		goto failed;

	int s = sizeof(WAVEFORMATEXTENSIBLE)+3*sizeof(size_t)+PacketSizes[0]+PacketSizes[1]+PacketSizes[2];
	s = (s+7) & -8;
	if( !allocFormat( s ) )
		goto failed;

	// prepare WAVEFORMAT
	WAVEFORMATEXTENSIBLE *wfext = (WAVEFORMATEXTENSIBLE *)getWaveFormat();

    wfext->Format.wFormatTag = 0xFFFE;		// WAVE_FORMAT_EXTENSIBLE
    wfext->Format.nChannels = vi.channels;
    wfext->Format.nSamplesPerSec = vi.rate;
    wfext->Format.nAvgBytesPerSec = 176400;
    wfext->Format.nBlockAlign = 20000;		// ahem, testing...
    wfext->Format.wBitsPerSample = 0;
    wfext->Format.cbSize = sizeof(WAVEFORMATEXTENSIBLE)
								+3*sizeof(size_t)
								+PacketSizes[0]+PacketSizes[1]+PacketSizes[2]
								-sizeof(WAVEFORMATEX);

//  wfext->Samples.wValidBitsPerSample = 0;       /* bits of precision  */
//  wfext->Samples.wSamplesPerBlock = 0;          /* valid if wBitsPerSample==0 */
    wfext->Samples.wReserved = 0;                 /* If neither applies, set to zero. */

	wfext->dwChannelMask = (1<<vi.channels)-1;

// {6BA47966-3F83-4178-9665-00F0BF6292E5}
// DEFINE_GUID(MEDIASUBTYPE_VorbisStream,0x6ba47966, 0x3f83, 0x4178, 0x96, 0x65, 0x0, 0xf0, 0xbf, 0x62, 0x92, 0xe5);
	GUID guid = { 0x6ba47966, 0x3f83, 0x4178, { 0x96, 0x65, 0x0, 0xf0, 0xbf, 0x62, 0x92, 0xe5 } };
	wfext->SubFormat = guid;

// append the Ogg headers to the Wav header
	char* p = (char*)getWaveFormat()+sizeof(WAVEFORMATEXTENSIBLE);
	*(size_t*)p = PacketSizes[0]; p += sizeof(size_t);
	*(size_t*)p = PacketSizes[1]; p += sizeof(size_t);
	*(size_t*)p = PacketSizes[2]; p += sizeof(size_t);
	memcpy( p, PacketAddresses[0], PacketSizes[0] ); p += PacketSizes[0];
	memcpy( p, PacketAddresses[1], PacketSizes[1] ); p += PacketSizes[1];
	memcpy( p, PacketAddresses[2], PacketSizes[2] ); p += PacketSizes[2];


// We need to know the length of the stream. Sure, there are other methods...

	int tot_pcm_samples = 0;
	
//	vorbis_synthesis_init( &vd, &vi );
//	vorbis_block_init( &vd, &vb );
	while( !eos ) {
		while( !eos ) {
			int result = ogg_sync_pageout( &oy, &og );
			if( result==0 ) break;
			if( result<0 ){
				//fprintf(stderr,"Corrupt or missing data in bitstream; continuing...\n");
			}else{
				ogg_stream_pagein( &os, &og );
				while( TRUE ) {
					result = ogg_stream_packetout( &os, &op );
					if( result==0 ) break;
					if( result<0 ) {
						/* no reason to complain; already complained above */
					}else{
						int samples;
						float **pcm;
						
						if( vorbis_synthesis( &vb, &op )==0 ) /* test for success! */
							vorbis_synthesis_blockin( &vd, &vb );

						if( (samples=vorbis_synthesis_pcmout( &vd, &pcm ))>0) {

							tot_pcm_samples += samples;
							
							vorbis_synthesis_read( &vd, samples );
						}	    
					}
				}
				if( ogg_page_eos( &og ) ) eos = 1;
			}
		}
		if( !eos ) {
			buffer = ogg_sync_buffer( &oy, 1 );
			bytes = fread( buffer, 1, 1, f);
			ogg_sync_wrote( &oy, bytes );
			if( bytes==0 ) eos = 1;
		}
	}
	
	ogg_stream_clear( &os );

	vorbis_block_clear( &vb );
	vorbis_dsp_clear( &vd );
	vorbis_comment_clear( &vc );
	vorbis_info_clear( &vi );


	// init some variables
	lSampleFirst	= 0;
	lSampleLast		= (tot_pcm_samples/20000)+1;

	pcm_samples = 0;
	pcm_written = 0;

	// reset the input

	fseek( f, 0, SEEK_SET );

	streamInit();
	while( pcm_samples<20000 ) {
		readPage();
		decodePage();
	}

	lCurrentSample = 0;

	if( Buffer ) free( Buffer );
	if( Header ) free( Header );
	if( Comments ) free( Comments );
	if( Codebook ) free( Codebook );

	// prepare StreamInfo
	streamInfo.fccType					= streamtypeAUDIO;
	streamInfo.fccHandler				= 0;
	streamInfo.dwFlags					= 0;
	streamInfo.wPriority				= 0;
	streamInfo.wLanguage				= 0;
	streamInfo.dwInitialFrames			= 0;
	streamInfo.dwScale					= 20000;
	streamInfo.dwRate					= vi.rate;
	streamInfo.dwStart					= 0;
	streamInfo.dwLength					= lSampleLast;
	streamInfo.dwSuggestedBufferSize	= 0;
	streamInfo.dwQuality				= 0xffffffff;
	streamInfo.dwSampleSize				= 0;

	return TRUE;
}


void AudioSourceOggVorbis::readPage() 
{
	char	*buffer;
	int		bytes;

	while( eos==0 && ogg_sync_pageout( &oy, &og )!=1 ) {
		buffer = ogg_sync_buffer( &oy, 1 );
		bytes = fread( buffer, 1, 1, f );
		ogg_sync_wrote( &oy, bytes );
		if( bytes==0 ) eos = 1;
	}
}


void AudioSourceOggVorbis::decodePage() 
{
	ogg_stream_pagein( &os, &og );
	while( TRUE ) {
		int result = ogg_stream_packetout( &os, &op );
		if( result==0 ) break;
		if( result<0 ) {
			/* no reason to complain; already complained above */
		}else{
			int samples;
			float **pcm;
			
			if( vorbis_synthesis( &vb, &op )==0 ) /* test for success! */
				vorbis_synthesis_blockin( &vd, &vb );

			if( (samples=vorbis_synthesis_pcmout( &vd, &pcm ))>0) {

				pcm_samples += samples;
							
				vorbis_synthesis_read( &vd, samples );
			}	    
		}
	}
	if( ogg_page_eos( &og ) ) eos = 1;
}


int AudioSourceOggVorbis::_read(LONG lStart, LONG lCount, LPVOID buffer, LONG cbBuffer, LONG *lBytesRead, LONG *lSamplesRead)
{
	*lSamplesRead = 0;
	*lBytesRead = 0;

	if( buffer ) {

		if (lStart != lCurrentSample) {

			ogg_stream_clear( &os );
			vorbis_block_clear( &vb );
			vorbis_dsp_clear( &vd );
			vorbis_comment_clear( &vc );
			vorbis_info_clear( &vi );
			fseek( f, 0, SEEK_SET );
			streamInit();

			readPage();
			lCurrentSample = 0;

			while( lStart != lCurrentSample ) {
				fgetpos( f, &pos );
				readPage();

				lCurrentSample++;
			}
		}

		__int64 pos2;
		fgetpos( f, &pos2 );
		uint32 frame_size = (uint32)pos2 - pos;
		if( frame_size>cbBuffer ) return AVIERR_OK;

		fseek( f, pos, SEEK_SET );
		fread( buffer, frame_size, 1, f );

		cbBuffer -= frame_size;
		buffer = (void *)(((unsigned long)buffer)+frame_size);
		(*lBytesRead) += frame_size;

		pcm_samples -= 20000;

		pos = pos2;
		while( pcm_samples<20000 ) {
			readPage();
			decodePage();
		}

		lCurrentSample++;
		(*lSamplesRead) = 1;
	}

	return AVIERR_OK;
}

///////////////////////////////////////////////////////////////////////////////////////////////////////

AudioSourceAC3::AudioSourceAC3(char *szFile, LONG inputBufferSize) 
{
	ac3File = fopen(szFile,"rb");
}


AudioSourceAC3::~AudioSourceAC3() 
{
	if (ac3File!=NULL) fclose(ac3File);
}

BOOL AudioSourceAC3::init() 
{
	WAVEFORMATEX *fmt, ac3WFmt;

	if( ac3File==NULL ) return FALSE;

	// extract WAVEFORMATEX from the AC3 file

	AC3FileSrc *ac3Src = new AC3FileSrc(ac3File);

	if( !ac3Src->Parse(&ac3WFmt) ) {
		delete ac3Src;
		return FALSE;
	}

	delete ac3Src;

	// allocate format structure
	if (!(fmt=(WAVEFORMATEX *) allocFormat(sizeof(WAVEFORMATEX)))) return FALSE;

	*fmt = ac3WFmt;
	{
		char szBuf[256];
		sprintf(szBuf, "FmtTag: 0x%x, SampFreq: %d, Channels: %d, bitrate: %d kb/s",
		fmt->wFormatTag,
		fmt->nSamplesPerSec,
		fmt->nChannels,
		(fmt->nAvgBytesPerSec*8)/1000);
		MessageBox(NULL,szBuf,"AC3 file parameters",MB_OK);
	}

	// get the length of the file
	fseek(ac3File,0,SEEK_END); 
	chunkDATA.cksize = ftell(ac3File);
	chunkDATA.dwDataOffset = 0;
	fseek(ac3File,0,SEEK_SET);

	bytesPerSample= getWaveFormat()->nBlockAlign; //getWaveFormat()->nAvgBytesPerSec / getWaveFormat()->nSamplesPerSec;
	lSampleFirst	= 0;
	lSampleLast	= chunkDATA.cksize / bytesPerSample;
	lCurrentSample= 0;

	streamInfo.fccType				= streamtypeAUDIO;
	streamInfo.fccHandler			= 0;
	streamInfo.dwFlags				= 0;
	streamInfo.wPriority			= 0;
	streamInfo.wLanguage			= 0;
	streamInfo.dwInitialFrames		= 0;
	streamInfo.dwScale				= bytesPerSample;
	streamInfo.dwRate				= getWaveFormat()->nAvgBytesPerSec;
	streamInfo.dwStart				= 0;
	streamInfo.dwLength				= chunkDATA.cksize / bytesPerSample;
	streamInfo.dwSuggestedBufferSize	= 0;
	streamInfo.dwQuality			= 0xffffffff;
	streamInfo.dwSampleSize			= bytesPerSample;

	return TRUE;
}

int AudioSourceAC3::_read(LONG lStart, LONG lCount, LPVOID buffer, LONG cbBuffer, LONG *lBytesRead, LONG *lSamplesRead) 
{
	LONG lBytes = lCount * bytesPerSample;

	if (lStart != lCurrentSample)
		if (-1 == fseek(ac3File, chunkDATA.dwDataOffset + bytesPerSample*lStart, SEEK_SET))
			return AVIERR_FILEREAD;

	if (lBytes != fread((char *)buffer, 1, lBytes, ac3File))
		return AVIERR_FILEREAD;

	*lSamplesRead = lCount;
	*lBytesRead = lBytes;

	lCurrentSample = lStart + lCount;

	return AVIERR_OK;
}