Commit 4020bc58 authored by norvell's avatar norvell
Browse files

Merge branch 'ericsson/td-renderer-with-radius-and-orientation' into 'main' 

[Non BE][Rend Non BE] 17 - TD Object renderer with radius and orientation

See merge request !499
parents f06bf0be a6d6cf44

apps/decoder.c

+19 −0
Original line number Diff line number Diff line
@@ -398,9 +398,17 @@ int main( 
        IVAS_RENDER_CONFIG_DATA renderConfig;



        /* sanity check */

#ifdef TD5

        if ( arg.outputFormat != IVAS_DEC_OUTPUT_BINAURAL_ROOM && arg.outputFormat != IVAS_DEC_OUTPUT_BINAURAL )

#else

        if ( arg.outputFormat != IVAS_DEC_OUTPUT_BINAURAL_ROOM )

#endif

        {

#ifdef TD5

            fprintf( stderr, "\nExternal Renderer Config is supported only for BINAURAL and BINAURAL_ROOM. Exiting. \n\n" );

#else

            fprintf( stderr, "\nExternal Renderer Config is supported only when BINAURAL_ROOM is used as output. Exiting. \n\n" );

#endif

            goto cleanup;

        }
@@ -1334,6 +1342,9 @@ static ivas_error decodeG192( 
    ivas_error error = IVAS_ERR_UNKNOWN;

    uint16_t numObj = 0;

    IVAS_DEC_BS_FORMAT bsFormat = IVAS_DEC_BS_UNKOWN;

#ifdef TD5

    IVAS_POSITION Pos[IVAS_MAX_PARAM_SPATIAL_SUBFRAMES];

#endif



    IsmFileWriter *ismWriters[IVAS_MAX_NUM_OBJECTS];

    for ( i = 0; i < IVAS_MAX_NUM_OBJECTS; ++i )
@@ -1408,13 +1419,21 @@ static ivas_error decodeG192( 
        {

            IVAS_QUATERNION Quaternions[IVAS_MAX_PARAM_SPATIAL_SUBFRAMES];



#ifdef TD5

            if ( ( error = HeadRotationFileReading( headRotReader, Quaternions, Pos ) ) != IVAS_ERR_OK )

#else

            if ( ( error = HeadRotationFileReading( headRotReader, Quaternions, frame ) ) != IVAS_ERR_OK )

#endif

            {

                fprintf( stderr, "\nError %s while reading head orientation from %s\n", IVAS_DEC_GetErrorMessage( error ), HeadRotationFileReader_getFilePath( headRotReader ) );

                goto cleanup;

            }



#ifdef TD5

            if ( ( error = IVAS_DEC_FeedHeadTrackData( hIvasDec, Quaternions, Pos ) ) != IVAS_ERR_OK )

#else

            if ( ( error = IVAS_DEC_FeedHeadTrackData( hIvasDec, Quaternions ) ) != IVAS_ERR_OK )

#endif

            {

                fprintf( stderr, "\nIVAS_DEC_FeedHeadTrackData failed: %s\n", IVAS_DEC_GetErrorMessage( error ) );

                goto cleanup;

apps/encoder.c

+23 −0
Original line number Diff line number Diff line
@@ -126,6 +126,9 @@ typedef struct 
#endif

#endif

    bool pca;

#ifdef TD5

    bool ism_extended_metadata;

#endif



} EncArguments;
@@ -378,7 +381,11 @@ int main( 
            }

            break;

        case IVAS_ENC_INPUT_ISM:

#ifdef TD5

            if ( ( error = IVAS_ENC_ConfigureForObjects( hIvasEnc, arg.inputFs, totalBitrate, arg.max_bwidth_user, bandwidth, arg.dtxConfig, arg.inputFormatConfig.ism.numObjects, arg.ism_extended_metadata ) ) != IVAS_ERR_OK )

#else

            if ( ( error = IVAS_ENC_ConfigureForObjects( hIvasEnc, arg.inputFs, totalBitrate, arg.max_bwidth_user, bandwidth, arg.dtxConfig, arg.inputFormatConfig.ism.numObjects ) ) != IVAS_ERR_OK )

#endif

            {

                fprintf( stderr, "\nIVAS_ENC_ConfigureForObjects failed: %s\n\n", IVAS_ENC_GetErrorMessage( error ) );

                goto cleanup;
@@ -876,6 +883,9 @@ static void initArgStruct( EncArguments *arg ) 
    arg->caConfig = IVAS_ENC_GetDefaultChannelAwareConfig();

    arg->ca_config_file = NULL;

    arg->mimeOutput = false;

#ifdef TD5

    arg->ism_extended_metadata = false;

#endif



#ifdef DEBUGGING

    arg->forcedMode = IVAS_ENC_FORCE_UNFORCED;
@@ -1263,6 +1273,13 @@ static bool parseCmdlIVAS_enc( 


            if ( i < argc - 4 )

            {

#ifdef TD5

                if ( argv[i][0] == '+' )

                {

                    argv[i]++;

                    arg->ism_extended_metadata = true;

                }

#endif

                if ( !is_digits_only( argv[i] ) )

                {

                    fprintf( stderr, "Error: Number of ISM channels must be an integer number!\n\n" );
@@ -1643,8 +1660,14 @@ static void usage_enc( void ) 
    fprintf( stdout, "EVS mono is default, for IVAS choose one of the following: -stereo, -ism, -sba, -masa, -mc\n" );

    fprintf( stdout, "-stereo [Mode]      : Stereo format, default is unified stereo \n" );

    fprintf( stdout, "                      optional for Mode: 1: DFT Stereo, 2: TD Stereo, 3: MDCT Stereo\n" );

#ifdef TD5

    fprintf( stdout, "-ism (+)Ch Files    : ISM format \n" );

    fprintf( stdout, "                      where Ch specifies the number of ISMs (1-4)\n" );

    fprintf( stdout, "                      where positive (+) means extended metadata format is used (including orientation and radius) \n" );

#else

    fprintf( stdout, "-ism Channels Files : ISM format \n" );

    fprintf( stdout, "                      where Channels specifies the number of ISMs (1-4)\n" );

#endif

    fprintf( stdout, "                      and Files specify input files containing metadata, one file per object\n" );

    fprintf( stdout, "                      (use NULL for no input metadata)\n" );

    fprintf( stdout, "-sba +/-Order       : Scene Based Audio input format (Ambisonics ACN/SN3D),\n" );

apps/renderer.c

+93 −23
Original line number Diff line number Diff line
@@ -517,6 +517,9 @@ int main( 
    int32_t delayTimeScale = 0;

    int16_t i, numChannels;

    ivas_error error = IVAS_ERR_OK;

#ifdef TD5

    IVAS_POSITION Pos[RENDERER_HEAD_POSITIONS_PER_FRAME];

#endif



#ifdef WMOPS

    reset_wmops();
@@ -622,19 +625,31 @@ int main( 
    }



    /* === Configure === */

#ifdef TD5

    if ( ( error = IVAS_REND_InitConfig( hIvasRend, ( args.outConfig.audioConfig == IVAS_REND_AUDIO_CONFIG_BINAURAL_ROOM ) || ( args.outConfig.audioConfig == IVAS_REND_AUDIO_CONFIG_BINAURAL ) ) ) != IVAS_ERR_OK )

#else

    if ( ( error = IVAS_REND_InitConfig( hIvasRend, strlen( args.renderConfigFilePath ) != 0 ) ) != IVAS_ERR_OK )

    {

        exit( -1 );

    }

#endif



        if ( args.renderConfigFilePath[0] != '\0' )

        {

            IVAS_RENDER_CONFIG_DATA renderConfig;



            /* sanity check */

#ifdef TD5

            if ( ( args.outConfig.audioConfig != IVAS_REND_AUDIO_CONFIG_BINAURAL_ROOM ) && ( args.outConfig.audioConfig != IVAS_REND_AUDIO_CONFIG_BINAURAL ) )

#else

        if ( args.outConfig.audioConfig != IVAS_REND_AUDIO_CONFIG_BINAURAL_ROOM )

#endif

            {

#ifdef TD5

                fprintf( stderr, "\nExternal Renderer Config is supported only when BINAURAL or BINAURAL_ROOM is used as output. Exiting. \n" );

#else

            fprintf( stderr, "\nExternal Renderer Config is supported only when BINAURAL_ROOM is used as output. Exiting. \n" );

#endif

                exit( -1 );

            }
@@ -860,12 +875,21 @@ int main( 
        if ( headRotReader != NULL )

        {

            IVAS_QUATERNION quatBuffer[RENDERER_HEAD_POSITIONS_PER_FRAME];

#ifdef TD5

            HeadRotationFileReading( headRotReader, quatBuffer, Pos );

            IVAS_REND_SetHeadRotation( hIvasRend, quatBuffer, Pos );

#else

            HeadRotationFileReading( headRotReader, quatBuffer, frame );

            IVAS_REND_SetHeadRotation( hIvasRend, quatBuffer );

#endif

        }

        else

        {

#ifdef TD5

            IVAS_REND_SetHeadRotation( hIvasRend, NULL, NULL );

#else

            IVAS_REND_SetHeadRotation( hIvasRend, NULL );

#endif

        }



        for ( i = 0; i < args.inConfig.numMultiChannelBuses; ++i )
@@ -1687,6 +1711,11 @@ void getMetadataFromFileReader( 


    objectMetadataBuffer->positions[objIdx].azimuth = ismMetadata.azimuth;

    objectMetadataBuffer->positions[objIdx].elevation = ismMetadata.elevation;

#ifdef TD5

    objectMetadataBuffer->positions[objIdx].radius = ismMetadata.radius;

    objectMetadataBuffer->positions[objIdx].yaw = ismMetadata.yaw;

    objectMetadataBuffer->positions[objIdx].pitch = ismMetadata.pitch;

#endif



    return;

}
@@ -1741,6 +1770,11 @@ static void IsmPositionProvider_getNextFrame( 
        {

            objectMetadataBuffer->positions[objIdx].azimuth = 0.0f;

            objectMetadataBuffer->positions[objIdx].elevation = 0.0f;

#ifdef TD5

            objectMetadataBuffer->positions[objIdx].radius = 1.0f;

            objectMetadataBuffer->positions[objIdx].yaw = 0.0f;

            objectMetadataBuffer->positions[objIdx].pitch = 0.0f;

#endif

        }



        /* Wrap azimuth to lie within (-180, 180] range */
@@ -1755,6 +1789,20 @@ static void IsmPositionProvider_getNextFrame( 


        /* Clamp elevation to lie within [-90, 90] range (can't be wrapped easily) */

        objectMetadataBuffer->positions[objIdx].elevation = min( max( objectMetadataBuffer->positions[objIdx].elevation, -90 ), 90 );

#ifdef TD5

        /* Wrap yaw to lie within (-180, 180] range */

        while ( objectMetadataBuffer->positions[objIdx].yaw < 0.0f )

        {

            objectMetadataBuffer->positions[objIdx].yaw += 360.0f;

        }

        while ( objectMetadataBuffer->positions[objIdx].yaw >= 360.0f )

        {

            objectMetadataBuffer->positions[objIdx].yaw -= 360.0f;

        }



        /* Clamp pitch to lie within [-90, 90] range (can't be wrapped easily) */

        objectMetadataBuffer->positions[objIdx].pitch = min( max( objectMetadataBuffer->positions[objIdx].pitch, -90 ), 90 );

#endif

    }



    ++positionProvider->frameCounter;
@@ -1989,6 +2037,28 @@ static void parseObjectPosition( 
    uint16_t *positionDuration )

{

    char *endptr;

#ifdef TD5

    int16_t read_values;

    float meta_prm[7] = { 0.0f, 0.0f, 1.0f, 0.0f, 1.0f, 0.0f, 0.0f };



    readNextMetadataChunk( line, "," );

    *positionDuration = (uint16_t) strtol( line, &endptr, 10 );

    readNextMetadataChunk( line, "\n" );



    read_values = (int16_t) sscanf( line, "%f,%f,%f,%f,%f,%f,%f", &meta_prm[0], &meta_prm[1], &meta_prm[2], &meta_prm[3], &meta_prm[4], &meta_prm[5], &meta_prm[6] );



    if ( read_values < 2 )

    {

        fprintf( stderr, "Error reading metadata\n" );

        exit( -1 );

    }



    position->azimuth = meta_prm[0];

    position->elevation = meta_prm[1];

    position->radius = meta_prm[2];

    position->yaw = meta_prm[5];

    position->pitch = meta_prm[6];

#else



    readNextMetadataChunk( line, "," );

    *positionDuration = (uint16_t) strtol( line, &endptr, 10 );
@@ -2015,7 +2085,7 @@ static void parseObjectPosition( 
        fprintf( stderr, "Error reading metadata\n" );

        exit( -1 );

    }



#endif

    return;

}

lib_com/common_api_types.h

+20 −0
Original line number Diff line number Diff line
@@ -78,6 +78,10 @@ typedef struct _IVAS_ISM_METADATA 
    float radius;

    float spread;

    float gainFactor;

#ifdef TD5

    float yaw;

    float pitch;

#endif

} IVAS_ISM_METADATA;



typedef struct
@@ -86,6 +90,14 @@ typedef struct 


} IVAS_QUATERNION;



#ifdef TD5

typedef struct

{

    float x, y, z;



} IVAS_POSITION;

#endif



typedef struct ivas_masa_metadata_frame_struct *IVAS_MASA_METADATA_HANDLE;

#ifdef FIX_350_MASA_DELAY_COMP

typedef struct ivas_masa_decoder_ext_out_meta_struct *MASA_DECODER_EXT_OUT_META_HANDLE;
@@ -111,6 +123,11 @@ typedef struct 
{

    float azimuth;

    float elevation;

#ifdef TD5

    float radius;

    float yaw;

    float pitch;

#endif

} IVAS_REND_AudioObjectPosition;



typedef struct _IVAS_ROOM_ACOUSTICS_CONFIG
@@ -132,6 +149,9 @@ typedef struct _IVAS_RENDER_CONFIG 
    IVAS_RENDER_TYPE_OVERRIDE renderer_type_override;

#endif

    IVAS_ROOM_ACOUSTICS_CONFIG_DATA room_acoustics;

#ifdef TD5

    float directivity[3];

#endif

} IVAS_RENDER_CONFIG_DATA, *IVAS_RENDER_CONFIG_HANDLE;



typedef struct _IVAS_LS_CUSTOM_LAYOUT

lib_com/ivas_cnst.h

+17 −0
Original line number Diff line number Diff line
@@ -320,6 +320,14 @@ typedef enum 
#define ISM_Q_STEP                              2.5f

#define ISM_Q_STEP_BORDER                       5.0f



#ifdef TD5

#define ISM_RADIUS_NBITS                       6 

#define ISM_RADIUS_MIN                         0.0f

#define ISM_RADIUS_DELTA                       0.25f /* Max radius = (2^ISM_RADIUS_NBITS-1)*0.25 = 15.75 */

#define ISM_EXTENDED_METADATA_BRATE            IVAS_64k

#define ISM_EXTENDED_METADATA_BITS             1

#endif 



/* Parametric ISM */

#define MAX_PARAM_ISM_NBANDS                    11

#define MAX_PARAM_ISM_NBANDS_WB                 9
@@ -349,7 +357,12 @@ typedef enum 
enum

{

    IND_ISM_NUM_OBJECTS,

#ifdef TD5

    IND_ISM_EXTENDED_FLAG = IND_ISM_NUM_OBJECTS + MAX_NUM_OBJECTS,

    IND_ISM_METADATA_FLAG = IND_ISM_EXTENDED_FLAG + MAX_NUM_OBJECTS, /* EN2VE: Is this not supposed to be in the loop part below, since it is one per ISm? */

#else

    IND_ISM_METADATA_FLAG = IND_ISM_NUM_OBJECTS + MAX_NUM_OBJECTS,

#endif

    IND_ISM_VAD_FLAG = IND_ISM_METADATA_FLAG + MAX_NUM_OBJECTS,

    IND_ISM_SCE_ID_DTX = IND_ISM_VAD_FLAG + MAX_NUM_OBJECTS,

    IND_ISM_NOISY_SPEECH_FLAG,
@@ -361,6 +374,10 @@ enum 
    IND_ISM_AZIMUTH = TAG_ISM_LOOP_START,

    IND_ISM_ELEVATION_DIFF_FLAG = TAG_ISM_LOOP_START,

    IND_ISM_ELEVATION = TAG_ISM_LOOP_START,

#ifdef TD5

    IND_ISM_RADIUS_DIFF_FLAG = TAG_ISM_LOOP_START,

    IND_ISM_RADIUS = TAG_ISM_LOOP_START,

#endif 

    TAG_ISM_LOOP_END = TAG_ISM_LOOP_START + 100, /* IVAS_fmToDo: to be reviewed once the final metadata are defined */

    /* --------- end of loop for objects ----------- */