Merge origin/main (32c876e9) · Commits · IVAS Codec Public Collaboration / IVAS Codec

apps/decoder.c

+138 −0

Original line number	Diff line number	Diff line
		@@ -41,6 +41,10 @@
		#ifdef IVAS_RTPDUMP
		#include "ivas_rtp_file.h"
		#endif
		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		#include "ivas_cnst.h"
		#endif

		#include "jbm_file_writer.h"
		#include "hrtf_file_reader.h"
		#include "ls_custom_file_reader.h"
		@@ -155,6 +159,9 @@ typedef struct
		uint16_t directivityPatternId[IVAS_MAX_NUM_OBJECTS];
		bool objEditEnabled;
		char *objEditFileName;
		#ifdef FIX_1419_MONO_STEREO_UMX
		bool evsMode;
		#endif

		} DecArguments;

		@@ -253,6 +260,10 @@ int main(
		IVAS_RENDER_FRAMESIZE asked_frame_size;
		IVAS_DEC_HRTF_BINARY_WRAPPER hHrtfBinary;
		ObjectEditFileReader *objectEditFileReader = NULL;
		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		IVAS_ROOM_ACOUSTICS_CONFIG_DATA **pAE = NULL;
		uint32_t aeCount = 0;
		#endif
		#ifdef DEBUGGING
		int32_t noClipping;
		int32_t cnt_frames_limited;
		@@ -477,7 +488,11 @@ int main(
		------------------------------------------------------------------------------------------/

		asked_frame_size = arg.renderFramesize;
		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		uint16_t aeID = arg.aeSequence.count > 0 ? arg.aeSequence.pID[0] : IVAS_DEFAULT_AEID;
		#else
		uint16_t aeID = arg.aeSequence.count > 0 ? arg.aeSequence.pID[0] : 65535;
		#endif

		#ifdef IVAS_RTPDUMP
		arg.enableHeadRotation = arg.enableHeadRotation \|\| arg.outputConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED \|\| arg.outputConfig == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM;
		@@ -658,6 +673,47 @@ int main(
		fprintf( stderr, "Failed to read renderer configuration from file %s\n\n", arg.renderConfigFilename );
		goto cleanup;
		}
		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		aeCount = RenderConfigReader_getAcousticEnvironmentCount( renderConfigReader );
		if ( aeCount > 0 )
		{
		uint32_t n;

		pAE = malloc( aeCount * sizeof( IVAS_ROOM_ACOUSTICS_CONFIG_DATA * ) );

		if ( pAE == NULL )
		{
		fprintf( stderr, "\nError: cannot allocate memory for acoustic environment array\n\n" );
		goto cleanup;
		}

		for ( n = 0; n < aeCount; n++ )
		{
		pAE[n] = NULL;

		if ( NULL == ( pAE[n] = malloc( sizeof( IVAS_ROOM_ACOUSTICS_CONFIG_DATA ) ) ) )
		{
		fprintf( stderr, "\nError: cannot allocate memory for acoustic environment\n\n" );
		goto cleanup;
		}
		}

		if ( ( error = RenderConfigReader_getAcousticEnvironments( renderConfigReader, pAE ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "Error while getting acoustic environments\n\n" );
		goto cleanup;
		}

		for ( n = 0; n < aeCount; n++ )
		{
		if ( ( error = IVAS_DEC_AddAcousticEnvironment( hIvasDec, *pAE[n] ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "Failed to add acoustic environments\n\n" );
		goto cleanup;
		}
		}
		}
		#endif

		if ( ( error = RenderConfigReader_getDirectivity( renderConfigReader, arg.directivityPatternId, renderConfig.directivity ) ) != IVAS_ERR_OK )
		{
		@@ -696,7 +752,11 @@ int main(

		if ( arg.outputConfig == IVAS_AUDIO_CONFIG_BINAURAL_ROOM_REVERB )
		{
		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		if ( ( error = IVAS_DEC_GetAcousticEnvironment( hIvasDec, aeID, &renderConfig.roomAcoustics ) ) == IVAS_ERR_OK )
		#else
		if ( ( error = RenderConfigReader_getAcousticEnvironment( renderConfigReader, aeID, &renderConfig.roomAcoustics ) ) == IVAS_ERR_OK )
		#endif
		{
		if ( RenderConfigReader_checkValues( &renderConfig ) != IVAS_ERR_OK )
		{
		@@ -835,6 +895,19 @@ cleanup:

		free( pcmBuf );

		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		if ( pAE != NULL )
		{
		uint16_t n;

		for ( n = 0; n < aeCount; n++ )
		{
		free( pAE[n] );
		}

		free( pAE );
		}
		#endif
		if ( arg.aeSequence.count > 0 )
		{
		free( arg.aeSequence.pID );
		@@ -1002,6 +1075,9 @@ static bool parseCmdlIVAS_dec(
		arg->output_Fs = IVAS_MAX_SAMPLING_RATE;
		arg->outputConfig = IVAS_AUDIO_CONFIG_MONO;
		arg->decMode = IVAS_DEC_MODE_IVAS;
		#ifdef FIX_1419_MONO_STEREO_UMX
		arg->evsMode = false;
		#endif
		arg->quietModeEnabled = false;
		arg->delayCompensationEnabled = true;
		arg->voipMode = false;
		@@ -1555,6 +1631,14 @@ static bool parseCmdlIVAS_dec(
		}
		i++;
		}
		#ifdef FIX_1419_MONO_STEREO_UMX
		else if ( strcmp( argv_to_upper, "-EVS" ) == 0 )
		{
		arg->evsMode = true;
		arg->decMode = IVAS_DEC_MODE_EVS;
		i++;
		}
		#endif

		/-----------------------------------------------------------------
		* Option not recognized
		@@ -1605,6 +1689,14 @@ static bool parseCmdlIVAS_dec(
		usage_dec();
		return false;
		}
		#ifdef FIX_1419_MONO_STEREO_UMX
		else if ( arg->non_diegetic_pan_enabled && arg->outputConfig == IVAS_AUDIO_CONFIG_STEREO && arg->evsMode )
		{
		fprintf( stderr, "Error: Both non-diegetic panning and stereo output specified!\n\n" );
		usage_dec();
		return false;
		}
		#endif

		if ( arg->outputMdFilename != NULL && arg->outputConfig != IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM )
		{
		@@ -1697,6 +1789,9 @@ static bool parseCmdlIVAS_dec(
		static void usage_dec( void )
		{
		fprintf( stdout, "Usage for EVS: IVAS_dec.exe [Options] Fs bitstream_file output_file\n" );
		#ifdef FIX_1419_MONO_STEREO_UMX
		fprintf( stdout, " OR usage for IVAS (below) with -evs option and OutputConf\n" );
		#endif
		fprintf( stdout, "Usage for IVAS: IVAS_dec.exe [Options] OutputConf Fs bitstream_file output_file\n\n" );

		fprintf( stdout, "Mandatory parameters:\n" );
		@@ -1713,6 +1808,9 @@ static void usage_dec( void )

		fprintf( stdout, "Options:\n" );
		fprintf( stdout, "--------\n" );
		#ifdef FIX_1419_MONO_STEREO_UMX
		fprintf( stdout, "-evs : Specify that the supplied bitstream is an EVS bitstream\n" );
		#endif
		fprintf( stdout, "-VOIP : VoIP mode: RTP in G192\n" );
		fprintf( stdout, "-VOIP_hf_only=0 : VoIP mode: EVS RTP Payload Format hf_only=0 in rtpdump\n" );
		#ifdef IVAS_RTPDUMP
		@@ -4238,6 +4336,46 @@ static void do_object_editing(
		editableParameters->ism_metadata[obj_idx].elevation = readInfo->obj_ele[obj_idx];
		}
		}
		#ifdef FIX_1427_OBJ_EDITING_EXT_METADATA
		/* object direction editing only for diegetic objects */
		if ( readInfo->obj_yaw_edited[obj_idx] )
		{
		if ( readInfo->obj_yaw_relative[obj_idx] )
		{
		/* yaw: apply relative edit + wrap */
		editableParameters->ism_metadata[obj_idx].yaw = fmodf( editableParameters->ism_metadata[obj_idx].yaw + readInfo->obj_yaw[obj_idx] + 540.f, 360.f ) - 180.f;
		}
		else
		{
		editableParameters->ism_metadata[obj_idx].yaw = readInfo->obj_yaw[obj_idx];
		}
		}
		if ( readInfo->obj_pitch_edited[obj_idx] )
		{
		if ( readInfo->obj_pitch_relative[obj_idx] )
		{
		/* pitch: apply relative edit + saturation */
		editableParameters->ism_metadata[obj_idx].pitch = fmaxf( fminf( editableParameters->ism_metadata[obj_idx].pitch + readInfo->obj_pitch[obj_idx], 90.f ), -90.f );
		}
		else
		{
		editableParameters->ism_metadata[obj_idx].pitch = readInfo->obj_pitch[obj_idx];
		}
		}
		/* radius editing only for diegetic objects */
		if ( readInfo->obj_radius_edited[obj_idx] )
		{
		if ( readInfo->obj_radius_relative[obj_idx] )
		{
		/* radius: apply relative edit + saturation */
		editableParameters->ism_metadata[obj_idx].radius = fmaxf( fminf( editableParameters->ism_metadata[obj_idx].radius * readInfo->obj_radius[obj_idx], OBJ_EDIT_RADIUS_MAX ), 0.0f );
		}
		else
		{
		editableParameters->ism_metadata[obj_idx].radius = readInfo->obj_radius[obj_idx];
		}
		}
		#endif
		}

		/* gain editing for all objects */

lib_com/common_api_types.h

+3 −0

Original line number	Diff line number	Diff line
		@@ -312,6 +312,9 @@ typedef enum
		#endif
		typedef struct _IVAS_ROOM_ACOUSTICS_CONFIG
		{
		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		uint16_t aeID; /* Acoustic environment ID*/
		#endif
		int16_t nBands; /* Number of frequency bands for which reverb properties are provided, integer, range [2..256] */
		float pFc_input[IVAS_CLDFB_NO_CHANNELS_MAX]; /* Center frequencies for which following values are provided: */
		float pAcoustic_rt60[IVAS_CLDFB_NO_CHANNELS_MAX]; /* - The room's T60 per center frequency */

lib_com/ivas_cnst.h

+13 −0

Original line number	Diff line number	Diff line
		@@ -177,6 +177,12 @@ typedef enum
		#define MAX_SPAR_INTERNAL_CHANNELS IVAS_SPAR_MAX_CH
		#define MAX_CLDFB_DIGEST_CHANNELS 3 /* == maximum of ParamISM TCs and ParamMC TCs */

		#ifdef FIX_GAIN_EDIT_LIMITS
		#define EDIT_GAIN_MIN 0.0630957f /* Minimum allowed gain edit with parametric modes: -24 dB */
		#define EDIT_GAIN_MAX 3.9810719f /* Maximum allowed gain edit: +12 dB */
		#endif


		typedef enum
		{
		TC_BUFFER_MODE_RENDERER,
		@@ -1251,7 +1257,11 @@ typedef enum
		----------------------------------------------------------------------------------/

		#define MC_LS_SETUP_BITS 3 /* number of bits for writing the MC LS configuration */
		#ifdef FIX_1419_MONO_STEREO_UMX
		#define LS_SETUP_CONVERSION_NUM_MAPPINGS 41 /* number of mappings for LS setup conversion */
		#else
		#define LS_SETUP_CONVERSION_NUM_MAPPINGS 35 /* number of mappings for LS setup conversion */
		#endif

		typedef enum
		{
		@@ -1592,6 +1602,9 @@ typedef enum
		#define RV_LENGTH_NR_FC ( RV_FILTER_MAX_FFT_SIZE / 2 ) + 1
		#define RV_LENGTH_NR_FC_16KHZ ( RV_FILTER_MAX_FFT_SIZE / 4 ) + 1
		#define IVAS_REVERB_DEFAULT_N_BANDS 31
		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		#define IVAS_DEFAULT_AEID ( 65535 )
		#endif

		#define LR_IAC_LENGTH_NR_FC ( RV_LENGTH_NR_FC )
		#define LR_IAC_LENGTH_NR_FC_16KHZ ( RV_LENGTH_NR_FC_16KHZ )

lib_com/options.h

100755 → 100644

+5 −0

Original line number	Diff line number	Diff line
		@@ -161,7 +161,9 @@

		#define RTP_S4_251135_CR26253_0016_REV1 /* RTP Pack/Unpack API corresponding to CR 26253 */
		#define IVAS_RTPDUMP /* RTPDUMP writing and reading for IVAS payloads */
		#define IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT /* RTPDUMP acoustic environment */
		#define FIXED_RTP_SEQUENCE_NUM /* Remove random sequence number initialization */
		#define ISM_PI_DATA /* Add reading and packing/unpacking of ISM PI data */
		#define DECODER_FORMAT_SWITCHING /* Re-initialize the decoder when the format/subformat of the incoming stream is changed */

		/* ################### Start BE switches ################################# */
		@@ -172,6 +174,8 @@
		#define TMP_1342_WORKAROUND_DEC_FLUSH_BROKEN_IN_SR /* FhG: Temporary workaround for incorrect implementation of decoder flush with split rendering */
		#define FIX_1413_IGF_INIT_PRINTOUT /* FhG: use correct variable for IGF initiliazation */
		#define RENDERER_MD_SYNC_DELAY_TO_INTEGER /* FhG: change data type of metadata sync delay in ext renderer to int16_t for better BASOP portability (and nicer code) */
		#define FIX_1427_OBJ_EDITING_EXT_METADATA /* Eri: Add support for extended metadata in object editing */
		#define FIX_1419_MONO_STEREO_UMX /* FhG: fix for issue 1419 : support upmix to all output formats for mono and stereo */


		/* #################### End BE switches ################################## */
		@@ -194,6 +198,7 @@
		#define FIX_1330_JBM_MEMORY_FIX /* VA: basop issue: 2179 fix non-BE difference in FIX_1330_JBM_MEMORY */
		#define NONBE_1380_OMASA_BUILD_DIFF /* Nokia: Fix for issue #1380: Large differences in OMASA output between Debug and Release builds */
		#define NONBE_FIX_1426_STEREO_PANNING_BETWEEN_OPT_LEVEL /* Nokia: Adjustments in remaining stereo panning functions to make them BE between Debug and Release */
		#define FIX_GAIN_EDIT_LIMITS /* Harmonize gain edit limits for all opertation points. For all modes, limit to max +12dB. For parametric modes, limit to min -24dB. */


		#define FIX_1430_EVS_STEREO_DMX_CHANNEL_DISAPPEARING /* Orange: Fix for basop issue 2184 - to prevent one channel from becoming inaudible in the mono downmix output */

lib_dec/ivas_init_dec.c

+49 −1

Original line number	Diff line number	Diff line
		@@ -1425,7 +1425,15 @@ ivas_error ivas_init_decoder(

		if ( output_config == IVAS_AUDIO_CONFIG_EXTERNAL )
		{
		#ifdef FIX_1419_MONO_STEREO_UMX
		if ( st_ivas->ivas_format == MONO_FORMAT )
		{
		hDecoderConfig->nchan_out = 1;
		}
		else if ( st_ivas->ivas_format == STEREO_FORMAT )
		#else
		if ( st_ivas->ivas_format == STEREO_FORMAT )
		#endif
		{
		hDecoderConfig->nchan_out = CPE_CHANNELS;
		}
		@@ -1478,6 +1486,13 @@ ivas_error ivas_init_decoder(
		st_ivas->hOutSetup.output_config = st_ivas->intern_config;
		st_ivas->hOutSetup.nchan_out_woLFE = audioCfg2channels( st_ivas->intern_config );
		}
		#ifdef FIX_1419_MONO_STEREO_UMX

		if ( st_ivas->ivas_format == MONO_FORMAT \|\| st_ivas->ivas_format == STEREO_FORMAT )
		{
		st_ivas->transport_config = ( st_ivas->ivas_format == MONO_FORMAT ) ? IVAS_AUDIO_CONFIG_MONO : IVAS_AUDIO_CONFIG_STEREO;
		}
		#endif

		/* Only initialize transport setup if it is used */
		if ( st_ivas->transport_config != IVAS_AUDIO_CONFIG_INVALID )
		@@ -1560,7 +1575,9 @@ ivas_error ivas_init_decoder(
		else if ( st_ivas->ivas_format == STEREO_FORMAT )
		{
		st_ivas->nchan_transport = CPE_CHANNELS;
		#ifndef FIX_1419_MONO_STEREO_UMX /* already set now by renderer_select() */
		st_ivas->intern_config = IVAS_AUDIO_CONFIG_STEREO;
		#endif

		st_ivas->nSCE = 0;
		st_ivas->nCPE = 1; /* in stereo, there is always only one CPE */
		@@ -2454,7 +2471,8 @@ ivas_error ivas_init_decoder(
		}

		/-----------------------------------------------------------------
		* LFE handles for rendering after rendering to adjust LFE delay to filter delay
		* LFE handles for rendering after rendering to adjust LFE delay to
		* filter delay
		-----------------------------------------------------------------/

		if ( st_ivas->mc_mode == MC_MODE_MCT \|\| st_ivas->mc_mode == MC_MODE_PARAMUPMIX )
		@@ -2777,6 +2795,10 @@ void ivas_initialize_handles_dec(
		st_ivas->hRenderConfig = NULL;
		st_ivas->hExtOrientationData = NULL;
		st_ivas->hCombinedOrientationData = NULL;
		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		st_ivas->acousticEnvironmentsCount = 0;
		st_ivas->pAcousticEnvironments = NULL;
		#endif

		st_ivas->hSplitBinRend = NULL;
		for ( i = 0; i < MAX_HEAD_ROT_POSES - 1; ++i )
		@@ -3023,6 +3045,15 @@ void ivas_destroy_dec(
		st_ivas->p_output_f[i] = NULL;
		}

		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		/* Acoustic environments */
		if ( st_ivas->pAcousticEnvironments != NULL )
		{
		free( st_ivas->pAcousticEnvironments );
		st_ivas->pAcousticEnvironments = NULL;
		}
		#endif

		/* main IVAS handle */
		free( st_ivas );

		@@ -3286,6 +3317,7 @@ static ivas_error doSanityChecks_IVAS(
		return IVAS_ERROR( IVAS_ERR_INVALID_OUTPUT_FORMAT, "Error: Non-diegetic panning not supported in this IVAS format" );
		}

		#ifndef FIX_1419_MONO_STEREO_UMX /* we now support basically everything for stereo */
		/* Verify stereo output configuration */
		if ( st_ivas->ivas_format == STEREO_FORMAT )
		{
		@@ -3298,10 +3330,26 @@ static ivas_error doSanityChecks_IVAS(
		else
		{
		if ( output_config == IVAS_AUDIO_CONFIG_INVALID )
		#else /* exclude invalid configs instead of matching valid ones */
		if ( output_config == IVAS_AUDIO_CONFIG_INVALID \|\|
		output_config == IVAS_AUDIO_CONFIG_ISM1 \|\|
		output_config == IVAS_AUDIO_CONFIG_ISM2 \|\|
		output_config == IVAS_AUDIO_CONFIG_ISM3 \|\|
		output_config == IVAS_AUDIO_CONFIG_ISM4 \|\|
		output_config == IVAS_AUDIO_CONFIG_MASA1 \|\|
		output_config == IVAS_AUDIO_CONFIG_MASA2 )
		#endif
		{
		return IVAS_ERROR( IVAS_ERR_INVALID_OUTPUT_FORMAT, "Incorrect output configuration specified!" );
		}
		if ( ( st_ivas->ivas_format == MONO_FORMAT \|\| st_ivas->ivas_format == STEREO_FORMAT ) &&
		( output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED \|\| output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM ) )
		{
		return IVAS_ERROR( IVAS_ERR_INVALID_OUTPUT_FORMAT, "Incorrect output configuration specified!" );
		}
		#ifndef FIX_1419_MONO_STEREO_UMX
		}
		#endif

		if ( ( output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_CODED \|\| output_config == IVAS_AUDIO_CONFIG_BINAURAL_SPLIT_PCM ) && output_Fs != 48000 )
		{