Merge branch 'main' into 549-paramism-output-too-quiet (b904d3fe) · Commits · IVAS Codec Public Collaboration / IVAS Codec

apps/encoder.c

+10 −13

Original line number	Diff line number	Diff line
		@@ -1225,16 +1225,6 @@ 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++;
		@@ -1312,6 +1302,13 @@ static bool parseCmdlIVAS_enc(
		}
		#endif /* DEBUGGING */
		}
		#ifdef BINAURAL_AUDIO_CMDLINE
		else if ( strcmp( argv_to_upper, "-BINAURAL" ) == 0 )
		{
		arg->is_binaural = true;
		i++;
		}
		#endif
		else if ( strcmp( argv_to_upper, "-ISM" ) == 0 )
		{
		arg->inputFormat = IVAS_ENC_INPUT_ISM;
		@@ -1703,9 +1700,6 @@ 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" );
		@@ -1737,6 +1731,9 @@ static void usage_enc( void )
		fprintf( stdout, " alternatively, B can be a text file where each line contains \"nb_frames B\"\n" );
		fprintf( stdout, "-no_delay_cmp : Turn off delay compensation\n" );
		fprintf( stdout, "-stereo_dmx_evs : Activate stereo downmix function for EVS.\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, "-mime : Mime output bitstream file format\n" );
		fprintf( stdout, " The encoder produces TS26.445 Annex.2.6 Mime Storage Format, (not RFC4867 Mime Format).\n" );
		fprintf( stdout, " default output bitstream file format is G.192\n" );

lib_com/ivas_cnst.h

+14 −0

Original line number	Diff line number	Diff line
		@@ -336,7 +336,12 @@ typedef enum

		#define ISM_NB_BITS_METADATA_NOMINAL ( ( SCE_CORE_16k_LOW_LIMIT - ACELP_16k_LOW_LIMIT ) / FRAMES_PER_SEC ) /* nominal number of metadata bits - used for configuration of Core-Coder modules */

		#ifdef FIX_532_ISM_MD_INACTIVE
		#define ISM_METADATA_MD_FLAG_BITS 1 /* flag to distinguish between NULL metadata and low-rate ISM_NO_META class */
		#define ISM_METADATA_INACTIVE_FLAG_BITS 1 /* flag to signal whether MD are sent in low-rate inactive frame */
		#else
		#define ISM_METADATA_VAD_FLAG_BITS 1
		#endif
		#define ISM_METADATA_FLAG_BITS 2

		#define ISM_NO_META 0
		@@ -390,6 +395,9 @@ typedef enum
		#define ISM_Q_STEP_LOW (ISM_Q_STEP * 2)
		#define ISM_Q_STEP_BORDER_LOW (ISM_Q_STEP_BORDER * 2)

		#ifdef FIX_532_ISM_MD_INACTIVE
		#define BRATE_ISM_INACTIVE 2450 /* CoreCoder bitrate in ISM inactive frames */
		#endif

		typedef enum
		{
		@@ -406,8 +414,14 @@ enum
		IND_ISM_EXTENDED_FLAG = IND_ISM_NUM_OBJECTS + MAX_NUM_OBJECTS,
		IND_ISM_EXTENDED_NDP_FLAG,
		IND_ISM_METADATA_FLAG,
		#ifdef FIX_532_ISM_MD_INACTIVE
		IND_ISM_MD_NULL_FLAG = IND_ISM_METADATA_FLAG + MAX_NUM_OBJECTS,
		IND_ISM_MD_INACTIVE_FLAG = IND_ISM_MD_NULL_FLAG + MAX_NUM_OBJECTS,
		IND_ISM_NOISY_SPEECH_FLAG = IND_ISM_MD_INACTIVE_FLAG + MAX_NUM_OBJECTS,
		#else
		IND_ISM_VAD_FLAG = IND_ISM_METADATA_FLAG + MAX_NUM_OBJECTS,
		IND_ISM_NOISY_SPEECH_FLAG = IND_ISM_VAD_FLAG + MAX_NUM_OBJECTS,
		#endif
		IND_ISM_SCE_ID_DTX,
		IND_ISM_DTX_COH_SCA,

lib_com/ivas_fb_mixer.c

+2 −0

Original line number	Diff line number	Diff line
		@@ -49,7 +49,9 @@

		static void ivas_get_active_bins( const int16_t pActive_bins, const int16_t pActive_bins_abs, const int16_t pStart_offset, const int16_t pStart_offset_ab, const int32_t sampling_rate );
		static void ivas_get_ld_fb_resp( float ppIdeal_FRs_re, float ppIdeal_FRs_im, float ppNew_FRs_re, float ppNew_FRs_im, const int16_t pActive_bins, const int16_t pStart_offset, const int16_t num_bands, const int16_t delay, const int32_t sampling_rate );
		#ifndef FIX_383_CLEAN_UP
		static int16_t ivas_fb_mixer_get_band_diff_non48k( const int32_t sampling_rate, const float delay_ms );
		#endif
		static const float *ivas_get_cheby_ramp( const int16_t delay );
		static void ivas_get_hanning_win( const int16_t len, float *pH_win );
		static ivas_error ivas_filterbank_setup( IVAS_FB_MIXER_HANDLE hFbMixer, const int32_t sampling_rate );

lib_com/ivas_ism_com.c

+73 −5

Original line number	Diff line number	Diff line
		@@ -49,10 +49,14 @@
		* Local constants
		-----------------------------------------------------------------------/

		#ifdef FIX_532_ISM_MD_INACTIVE
		#define BITS_ISM_INACTIVE ( BRATE_ISM_INACTIVE / FRAMES_PER_SEC )
		#else
		#define FRMS_PER_SECOND ( 1000000000 / FRAME_SIZE_NS )

		#define BRATE_ISM_INACTIVE 2450 /* CoreCoder bitrate in ISM inactive frames */
		#define BITS_ISM_INACTIVE ( BRATE_ISM_INACTIVE / FRMS_PER_SECOND )
		#endif

		#define BETA_ISM_LOW_IMP 0.6f
		#define BETA_ISM_MEDIUM_IMP 0.8f
		@@ -74,7 +78,11 @@ static void bitbudget_to_brate(

		for ( i = 0; i < N; i++ )
		{
		#ifdef FIX_532_ISM_MD_INACTIVE
		y[i] = FRAMES_PER_SEC * x[i];
		#else
		y[i] = FRMS_PER_SECOND * x[i];
		#endif
		}

		return;
		@@ -93,7 +101,11 @@ ivas_error ivas_ism_config(
		const int16_t nchan_ism, /* i : number of objects */
		ISM_METADATA_HANDLE hIsmMeta[], /* i/o: ISM metadata handles */
		const int16_t ism_extended_metadata_flag, /* i : extended metadata flag */
		#ifdef FIX_532_ISM_MD_INACTIVE
		const int16_t null_metadata_flag[MAX_NUM_OBJECTS], /* i : NULL MD flag */
		#else
		const int16_t localVAD[MAX_NUM_OBJECTS], /* i : local VAD flag */
		#endif
		const int16_t ism_imp[], /* i : ISM importance flags */
		int32_t element_brate[], /* o : element bitrate per object */
		int32_t total_brate[], /* o : total bitrate per object */
		@@ -120,12 +132,20 @@ ivas_error ivas_ism_config(
		{
		for ( ch = 0; ch < n_ISms; ch++ )
		{
		#ifdef FIX_532_ISM_MD_INACTIVE
		ism_metadata_flag_global \|= ism_imp[ch];
		#else
		ism_metadata_flag_global \|= hIsmMeta[ch]->ism_metadata_flag;
		#endif
		}
		}

		/* decision about bitrates per channel - constant during the session (at one ivas_total_brate) */
		#ifdef FIX_532_ISM_MD_INACTIVE
		bits_ism = (int16_t) ( ism_total_brate / FRAMES_PER_SEC );
		#else
		bits_ism = (int16_t) ( ism_total_brate / FRMS_PER_SECOND );
		#endif
		set_s( bits_element, bits_ism / n_ISms, n_ISms );
		bits_element[n_ISms - 1] += bits_ism % n_ISms;
		bitbudget_to_brate( bits_element, element_brate, n_ISms );
		@@ -146,10 +166,26 @@ ivas_error ivas_ism_config(

		for ( ch = 0; ch < n_ISms; ch++ )
		{
		#ifdef FIX_532_ISM_MD_INACTIVE
		if ( null_metadata_flag[ch] )
		{
		nb_bits_metadata[0] += ISM_METADATA_MD_FLAG_BITS;
		nb_bits_metadata[0] += ISM_METADATA_FLAG_BITS;
		}
		else
		{
		if ( ism_imp[ch] == ISM_NO_META )
		{
		nb_bits_metadata[0] += ISM_METADATA_MD_FLAG_BITS;
		nb_bits_metadata[0] += ISM_METADATA_INACTIVE_FLAG_BITS;
		}
		}
		#else
		if ( hIsmMeta[ch]->ism_metadata_flag == 0 )
		{
		nb_bits_metadata[0] += ISM_METADATA_VAD_FLAG_BITS;
		}
		#endif
		}
		}

		@@ -175,7 +211,11 @@ ivas_error ivas_ism_config(
		diff = 0;
		for ( ch = 0; ch < n_ISms; ch++ )
		{
		#ifdef FIX_532_ISM_MD_INACTIVE
		if ( ism_imp[ch] == ISM_NO_META )
		#else
		if ( hIsmMeta[ch]->ism_metadata_flag == 0 && localVAD[ch] == 0 )
		#endif
		{
		diff += bits_CoreCoder[ch] - BITS_ISM_INACTIVE;
		bits_CoreCoder[ch] = BITS_ISM_INACTIVE;
		@@ -212,18 +252,34 @@ ivas_error ivas_ism_config(
		{
		int16_t limit;

		#ifdef FIX_532_ISM_MD_INACTIVE
		limit = MIN_BRATE_SWB_BWE / FRAMES_PER_SEC;
		#else
		limit = MIN_BRATE_SWB_BWE / FRMS_PER_SECOND;
		#endif
		if ( element_brate[ch] < MIN_BRATE_SWB_STEREO ) /* replicate function set_bw() -> check the coded audio band-width */
		{
		#ifdef FIX_532_ISM_MD_INACTIVE
		limit = MIN_BRATE_WB_BWE / FRAMES_PER_SEC;
		#else
		limit = MIN_BRATE_WB_BWE / FRMS_PER_SECOND;
		#endif
		}
		else if ( element_brate[ch] >= SCE_CORE_16k_LOW_LIMIT ) /* replicate function set_ACELP_flag() -> it is not intended to switch the ACELP internal sampling rate within an object */
		{
		/limit = SCE_CORE_16k_LOW_LIMIT;/
		#ifdef FIX_532_ISM_MD_INACTIVE
		limit = ( ACELP_16k_LOW_LIMIT + SWB_TBE_1k6 ) / FRAMES_PER_SEC;
		#else
		limit = ( ACELP_16k_LOW_LIMIT + SWB_TBE_1k6 ) / FRMS_PER_SECOND;
		#endif
		}

		#ifdef FIX_532_ISM_MD_INACTIVE
		if ( ism_imp[ch] == ISM_NO_META )
		#else
		if ( ism_imp[ch] == ISM_NO_META && localVAD[ch] == 0 )
		#endif
		{
		tmp = BITS_ISM_INACTIVE;
		}
		@@ -270,10 +326,18 @@ ivas_error ivas_ism_config(
		diff = 0;
		for ( ch = 0; ch < n_ISms; ch++ )
		{
		#ifdef FIX_532_ISM_MD_INACTIVE
		limit_high = IVAS_512k / FRAMES_PER_SEC;
		#else
		limit_high = IVAS_512k / FRMS_PER_SECOND;
		#endif
		if ( element_brate[ch] < SCE_CORE_16k_LOW_LIMIT ) /* replicate function set_ACELP_flag() -> it is not intended to switch the ACELP internal sampling rate within an object */
		{
		#ifdef FIX_532_ISM_MD_INACTIVE
		limit_high = ACELP_12k8_HIGH_LIMIT / FRAMES_PER_SEC;
		#else
		limit_high = ACELP_12k8_HIGH_LIMIT / FRMS_PER_SECOND;
		#endif
		}

		tmp = (int16_t) min( bits_CoreCoder[ch], limit_high );
		@@ -316,7 +380,11 @@ ivas_error ivas_ism_config(
		if ( nb_bits_metadata != NULL )
		{
		int32_t tmpL;
		#ifdef FIX_532_ISM_MD_INACTIVE
		tmpL = sum_l( total_brate, n_ISms ) + bits_side * FRAMES_PER_SEC;
		#else
		tmpL = sum_l( total_brate, n_ISms ) + bits_side * FRMS_PER_SECOND;
		#endif
		if ( sum_l( element_brate, n_ISms ) != tmpL )
		{
		return IVAS_ERROR( IVAS_ERR_INTERNAL_FATAL, "\nError: Mismatch in ISM bit-budget distribution. Exiting!\n" );

lib_com/ivas_prot.h

+4 −0

Original line number	Diff line number	Diff line
		@@ -873,7 +873,11 @@ ivas_error ivas_ism_config(
		const int16_t nchan_ism, /* i : number of objects */
		ISM_METADATA_HANDLE hIsmMeta[], /* i/o: ISM metadata handles */
		const int16_t ism_extended_metadata_flag, /* i : extended metadata flag */
		#ifdef FIX_532_ISM_MD_INACTIVE
		const int16_t null_metadata_flag[MAX_NUM_OBJECTS], /* i : NULL MD flag */
		#else
		const int16_t localVAD[MAX_NUM_OBJECTS], /* i : local VAD flag */
		#endif
		const int16_t ism_imp[], /* i : ISM importance flags */
		int32_t element_brate[], /* o : element bitrate per object */
		int32_t total_brate[], /* o : total bitrate per object */