Commit a6425dc8 authored by stoutjesdijk's avatar stoutjesdijk 🎧
Browse files

Merge remote-tracking branch 'origin/main' into 196-refactor-renderer-output-configuration-cleanup

parents 202543a3 830ab066

apps/encoder.c

+32 −0
Original line number Diff line number Diff line
@@ -98,6 +98,9 @@ typedef struct 
    char *outputBitstreamFilename;

    int32_t inputFs;

    IVAS_ENC_INPUT_FORMAT inputFormat;

#ifdef BINAURAL_AUDIO_CMDLINE

    bool is_binaural;

#endif

    EncInputFormatConfig inputFormatConfig;

    bool max_bwidth_user;

    IVAS_ENC_BANDWIDTH maxBandwidth;
@@ -362,17 +365,29 @@ int main( 
    switch ( arg.inputFormat )

    {

        case IVAS_ENC_INPUT_MONO:

#ifdef BINAURAL_AUDIO_CMDLINE

            if ( ( error = IVAS_ENC_ConfigureForMono( hIvasEnc, arg.inputFs, totalBitrate, arg.max_bwidth_user, bandwidth, arg.dtxConfig, caConfig, arg.inputFormatConfig.stereoToMonoDownmix, arg.is_binaural ) ) != IVAS_ERR_OK )

#else

            if ( ( error = IVAS_ENC_ConfigureForMono( hIvasEnc, arg.inputFs, totalBitrate, arg.max_bwidth_user, bandwidth, arg.dtxConfig, caConfig, arg.inputFormatConfig.stereoToMonoDownmix ) ) != IVAS_ERR_OK )

#endif

            {

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

                goto cleanup;

            }

            break;

        case IVAS_ENC_INPUT_STEREO:

#ifdef BINAURAL_AUDIO_CMDLINE

#ifdef DEBUGGING

            if ( ( error = IVAS_ENC_ConfigureForStereo( hIvasEnc, arg.inputFs, totalBitrate, arg.max_bwidth_user, bandwidth, arg.dtxConfig, arg.is_binaural, arg.inputFormatConfig.stereoMode ) ) != IVAS_ERR_OK )

#else

            if ( ( error = IVAS_ENC_ConfigureForStereo( hIvasEnc, arg.inputFs, totalBitrate, arg.max_bwidth_user, bandwidth, arg.dtxConfig, arg.is_binaural ) ) != IVAS_ERR_OK )

#endif

#else

#ifdef DEBUGGING

            if ( ( error = IVAS_ENC_ConfigureForStereo( hIvasEnc, arg.inputFs, totalBitrate, arg.max_bwidth_user, bandwidth, arg.dtxConfig, arg.inputFormatConfig.stereoMode ) ) != IVAS_ERR_OK )

#else

            if ( ( error = IVAS_ENC_ConfigureForStereo( hIvasEnc, arg.inputFs, totalBitrate, arg.max_bwidth_user, bandwidth, arg.dtxConfig ) ) != IVAS_ERR_OK )

#endif

#endif

            {

                fprintf( stderr, "\nIVAS_ENC_ConfigureForStereo failed: %s\n\n", IVAS_ENC_GetErrorMessage( error ) );
@@ -865,6 +880,9 @@ static void initArgStruct( EncArguments *arg ) 
    arg->outputBitstreamFilename = NULL;

    arg->inputFs = 0;

    arg->inputFormat = IVAS_ENC_INPUT_MONO;

#ifdef BINAURAL_AUDIO_CMDLINE

    arg->is_binaural = false;

#endif

    arg->inputFormatConfig.stereoToMonoDownmix = false;

    arg->max_bwidth_user = false;

    arg->maxBandwidth = IVAS_ENC_BANDWIDTH_UNDEFINED;
@@ -1207,6 +1225,16 @@ static bool parseCmdlIVAS_enc( 
         * IVAS Formats

         *-----------------------------------------------------------------*/



#ifdef BINAURAL_AUDIO_CMDLINE

        else if ( strcmp( argv_to_upper, "-BINAURAL" ) == 0 )

        {

            i++;

            if ( strcmp( argv_to_upper, "-BINAURAL" ) == 0 )

            {

                arg->is_binaural = true;

            }

        }

#endif

        else if ( strcmp( argv_to_upper, "-STEREO" ) == 0 )

        {

            i++;
@@ -1487,6 +1515,7 @@ static bool parseCmdlIVAS_enc( 
        {

            arg->inputFormat = IVAS_ENC_INPUT_MONO;

            arg->inputFormatConfig.stereoToMonoDownmix = true;



            i++;

        }

        else if ( strcmp( argv_to_upper, "-BYPASS" ) == 0 ) // TODO: should be renamed to "-pca"
@@ -1674,6 +1703,9 @@ static void usage_enc( void ) 
    fprintf( stdout, "Options:\n" );

    fprintf( stdout, "--------\n" );

    fprintf( stdout, "EVS mono is default, for IVAS choose one of the following: -stereo, -ism, -sba, -masa, -mc\n" );

#ifdef BINAURAL_AUDIO_CMDLINE

    fprintf( stdout, "-binaural           : Optional indication that input is binaural audio (to be used with -stereo or -stereo_dmx_evs)\n" );

#endif

    fprintf( stdout, "-stereo             : Stereo format \n" );

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

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

lib_com/ivas_cnst.h

+35 −0
Original line number Diff line number Diff line
@@ -1790,6 +1790,41 @@ typedef enum 
#define IVAS_LIMITER_THRESHOLD                  32729           /* -0.01 dBFS */

#define IVAS_LIMITER_ATTACK_SECONDS             0.005f



#ifdef ENHANCED_STEREO_DMX

/*----------------------------------------------------------------------------------*

 * Stereo downmix EVS constants

 *----------------------------------------------------------------------------------*/



#define STEREO_DMX_EVS_PHA_LEN_16 48

#define STEREO_DMX_EVS_FAD_LEN_16 160

#define STEREO_DMX_EVS_PHA_LEN_32 96

#define STEREO_DMX_EVS_FAD_LEN_32 320

#define STEREO_DMX_EVS_PHA_LEN_48 96

#define STEREO_DMX_EVS_FAD_LEN_48 480



#define STEREO_DMX_EVS_SUBBAND_SIZE 2

#define STEREO_DMX_EVS_NB_SUBBAND_MAX (L_FRAME48k / (2 * STEREO_DMX_EVS_SUBBAND_SIZE))



#define STEREO_DMX_EVS_PHA_LEN_MAX 96 /* Max of PHA_LEN */

#define STEREO_DMX_EVS_FAD_LEN_MAX 480 /* Max of FAD_LEN */



#define STEREO_DMX_EVS_DATA_LEN_MAX (STEREO_DMX_EVS_PHA_LEN_MAX + L_FRAME48k)



typedef enum

{

    STEREO_DMX_EVS_PHA_IPD,

    STEREO_DMX_EVS_PHA_IPD2,

    STEREO_DMX_EVS_NO_PHA



} STEREO_DMX_EVS_PHA;



typedef enum

{

    STEREO_DMX_EVS_PRC_POC,

    STEREO_DMX_EVS_PRC_PHA,



} STEREO_DMX_EVS_PRC;

#endif



#endif

/* clang-format on */

lib_com/ivas_prot.h

+5 −0
Original line number Diff line number Diff line
@@ -61,7 +61,12 @@ void stereo_dmx_evs_enc( 
    STEREO_DMX_EVS_ENC_HANDLE hStereoDmxEVS,                    /* i/o: Stereo downmix for EVS encoder handle   */

    const int32_t input_Fs,                                     /* i  : input sampling rate                     */

    int16_t data[CPE_CHANNELS * L_FRAME48k],                    /* i/o: input signal                            */

#ifdef BINAURAL_AUDIO_CMDLINE

    const int16_t n_samples,                                    /* i  : number of input samples                 */

    const bool is_binaural                                      /* i  : indication that input is binaural audio */

#else

    const int16_t n_samples                                     /* i  : number of input samples                 */

#endif

);



/*! r: number of channels to be analysed */

lib_com/options.h

100644 → 100755
+4 −0
Original line number Diff line number Diff line
@@ -234,6 +234,10 @@ 
#define FIX_TODO_NON_DIEGETIC_PAN_NOT_IMPLELENTED_IN_RENDERER /* ..\apps\renderer.c(240):  .... (todo: implementation)",*/

#define REMOVE_OBS_CODE                                 /* FhG: Remove unnecessary assignement after LFE cleanup (Issue #451)*/



#define ENHANCED_STEREO_DMX                             /* Orange : Contribution 48 - Enhanced stereo downmix. */



#define BINAURAL_AUDIO_CMDLINE



/* ################## End DEVELOPMENT switches ######################### */

/* clang-format on */

lib_enc/ivas_stat_enc.h

+44 −0
Original line number Diff line number Diff line
@@ -1009,9 +1009,50 @@ typedef struct stereo_dmx_evs_phase_only_correlation_structure 


} STEREO_DMX_EVS_POC_DATA, *STEREO_DMX_EVS_POC_HANDLE;



#ifdef ENHANCED_STEREO_DMX

typedef struct stereo_dmx_evs_correlation_filter_structure

{

    int16_t init_frmCntr;

    float isd_rate_s;

    float iccr_s;

    float ipd_ff[STEREO_DMX_EVS_NB_SUBBAND_MAX];

    float Pr[STEREO_DMX_EVS_NB_SUBBAND_MAX];

    float Pi[STEREO_DMX_EVS_NB_SUBBAND_MAX];

    float rfft_ipd_coef[L_FRAME48k / 2 + 1];



    int16_t pha_len;

    int16_t fad_len;



    float win[STEREO_DMX_EVS_PHA_LEN_MAX];

    float fad_g[STEREO_DMX_EVS_FAD_LEN_MAX];

    float *p_prev_taps[CPE_CHANNELS], prev_taps[CPE_CHANNELS][STEREO_DMX_EVS_PHA_LEN_MAX];

    float *p_curr_taps[CPE_CHANNELS], curr_taps[CPE_CHANNELS][STEREO_DMX_EVS_PHA_LEN_MAX];



    float data_mem[CPE_CHANNELS][STEREO_DMX_EVS_PHA_LEN_MAX];



    STEREO_DMX_EVS_PHA curr_pha;

    STEREO_DMX_EVS_PHA prev_pha;

    int16_t pha_hys_cnt;



    int16_t prc_thres;

    STEREO_DMX_EVS_PRC curr_prc;

    STEREO_DMX_EVS_PRC prev_prc;

    int16_t prc_hys_cnt;

    float fad_g_prc[L_FRAME48k];

    int16_t fad_len_prc;



    float trns_aux_energy[CPE_CHANNELS];

    float crst_fctr;



} STEREO_DMX_EVS_PHA_DATA, *STEREO_DMX_EVS_PHA_HANDLE;



#endif

typedef struct stereo_dmx_evs_enc_data_structure

{

    STEREO_DMX_EVS_POC_HANDLE hPOC;

#ifdef ENHANCED_STEREO_DMX

    STEREO_DMX_EVS_PHA_HANDLE hPHA;

#endif



    float itd;
@@ -1055,6 +1096,9 @@ typedef struct encoder_config_structure 
    int16_t nchan_inp;                              /* number of input audio channels */

    int16_t max_bwidth;                             /* maximum encoded bandwidth */

    IVAS_FORMAT ivas_format;                        /* IVAS format */

#ifdef BINAURAL_AUDIO_CMDLINE

	bool is_binaural;                               /* flag indicating if input is binaural audio */

#endif



    int16_t element_mode_init;                      /* element mode used at initialization */

    int16_t stereo_dmx_evs;                         /* flag to indicate that stereo downmix for EVS encoder */