Merge remote-tracking branch 'remotes/origin/main' into float-1579-remove-extra-encoder-app (0478af77) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_cnst.h

+5 −0

Original line number	Diff line number	Diff line
		@@ -1216,7 +1216,12 @@ typedef enum
		{
		MASA_STEREO_NOT_DEFINED,
		MASA_STEREO_SPACED_MICS,
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		MASA_STEREO_DOWNMIX,
		MASA_DUAL_MONO
		#else
		MASA_STEREO_DOWNMIX
		#endif
		} MASA_TRANSPORT_SIGNAL_TYPE;

		typedef enum

lib_com/options.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -182,6 +182,7 @@
		#define FIX_1452_DEFAULT_REVERB /* Nokia/Philips/FhG: Fix default room presets and their usage in renderer */
		#define FIX_FMSW_DEC /* float issue 1542: fix JBM issue in format switching */
		#define FIX_FMSW_DEC_2 /* float issue 1575: fix crash for format switching when bitsream starts with EVS */
		#define FIX_FLOAT_1578_OMASA_REND_SPIKES /* Nokia: Float issue 1578: Fix spikes and collapsed perception in OMASA/MASA rendering to FOA/HOA */

		/* ##################### End NON-BE switches ########################### */

lib_dec/ivas_dirac_dec.c

+4 −0

Original line number	Diff line number	Diff line
		@@ -274,7 +274,11 @@ static ivas_error ivas_dirac_rend_config(
		return ( IVAS_ERROR( IVAS_ERR_FAILED_ALLOC, "Can not allocate memory for DirAC\n" ) );
		}
		}
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		ivas_masa_init_stereotype_detection( hDirACRend->masa_stereo_type_detect, ivas_total_brate );
		#else
		ivas_masa_init_stereotype_detection( hDirACRend->masa_stereo_type_detect );
		#endif
		}
		else
		{

lib_rend/ivas_dirac_output_synthesis_dec.c

+7 −0

Original line number	Diff line number	Diff line
		@@ -1761,6 +1761,13 @@ void ivas_dirac_dec_compute_directional_responses(
		direct_response[1] = 1.0f;
		}
		}
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		else if ( transport_signal_type == MASA_DUAL_MONO )
		{
		direct_response[0] = 1.0f;
		/* direct_response[1] is not adjusted for dual mono input */
		}
		#endif
		else
		{
		set_f( direct_response, 1.0f, hDirACRend->num_protos_ambi );

lib_rend/ivas_dirac_rend.c

+227 −0

Original line number	Diff line number	Diff line
		@@ -45,6 +45,15 @@
		#endif
		#include "wmc_auto.h"

		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		/------------------------------------------------------------------------------------------
		* Local constants
		------------------------------------------------------------------------------------------/

		/* Constants for MASA dual mono detection */
		#define MASA_DUAL_MONO_TAU1 0.01562500000000000f /* 2^-6, which is about -18 dB */
		#define MASA_DUAL_MONO_TAU2 0.00000762939453125f /* 2^-17, which is about -51 dB */
		#endif

		/*-------------------------------------------------------------------------
		* ivas_dirac_allocate_parameters()
		@@ -1328,9 +1337,23 @@ void protoSignalComputation2(

		float interpolatorSpaced = 0.0f;
		float interpolatorDmx = 1.0f;
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		float interpolatorDualMono = 0.0f;
		int16_t max_band_diff_ene;
		#endif

		int16_t dipole_freq_range[2];
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		float tempSpaced, tempDmx, tempDualMono;

		max_band_diff_ene = IVAS_CLDFB_NO_CHANNELS_MAX;
		if ( stereo_type_detect != NULL )
		{
		max_band_diff_ene = min( stereo_type_detect->max_band_diff_ene, num_freq_bands ); /* Local version for correct maximum */
		}
		#else
		float tempSpaced, tempDmx;
		#endif

		if ( isloudspeaker )
		{
		@@ -1381,6 +1404,11 @@ void protoSignalComputation2(
		right_hi_power = 0.0f;
		total_hi_power = 0.0f;

		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		stereo_type_detect->inst_diff_bb_power = 0.0f;
		stereo_type_detect->inst_total_bb_power = 0.0f;

		#endif
		dipole_freq_range[0] = stereo_type_detect->dipole_freq_range[0];
		dipole_freq_range[1] = stereo_type_detect->dipole_freq_range[1];

		@@ -1395,11 +1423,27 @@ void protoSignalComputation2(
		{
		interpolatorSpaced = ( (float) ( stereo_type_detect->interpolator ) ) / ( (float) MASA_STEREO_INTERPOLATION_SLOTS );
		interpolatorDmx = 1.0f - interpolatorSpaced;
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		interpolatorDualMono = interpolatorDmx;
		#endif
		}
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		else if ( stereo_type_detect->type_change_direction == MASA_DUAL_MONO )
		{
		interpolatorDualMono = ( (float) ( stereo_type_detect->interpolator ) ) / ( (float) MASA_STEREO_INTERPOLATION_SLOTS );
		interpolatorDmx = 1.0f - interpolatorDualMono;
		interpolatorSpaced = interpolatorDmx;
		}
		else /* MASA_STEREO_DOWNMIX */
		#else
		else
		#endif
		{
		interpolatorDmx = ( (float) ( stereo_type_detect->interpolator ) ) / ( (float) MASA_STEREO_INTERPOLATION_SLOTS );
		interpolatorSpaced = 1.0f - interpolatorDmx;
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		interpolatorDualMono = interpolatorSpaced;
		#endif
		}
		}

		@@ -1439,18 +1483,39 @@ void protoSignalComputation2(
		sum_total_ratio[l] = stereo_type_detect->sum_power[l] / ( stereo_type_detect->total_power[l] + EPSILON );
		}

		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		if ( l < max_band_diff_ene )
		{
		RealSubtract = RealBuffer[0][0][l] - RealBuffer[1][0][l];
		ImagSubtract = ImagBuffer[0][0][l] - ImagBuffer[1][0][l];
		stereo_type_detect->inst_diff_bb_power += RealSubtract * RealSubtract + ImagSubtract * ImagSubtract;

		if ( l == max_band_diff_ene - 1 )
		{
		/* Stores sum of reference power (i.e., sum of transport channels) up to max_band_diff_ene - 1. */
		stereo_type_detect->inst_total_bb_power = total_bb_power; /* Total energy before smoothing */
		}

		if ( l == 0 )
		{
		stereo_type_detect->subtract_power_y += RealSubtract * RealSubtract + ImagSubtract * ImagSubtract;
		}
		}
		#else
		if ( l == 0 )
		{
		RealSubtract = RealBuffer[0][0][l] - RealBuffer[1][0][l];
		ImagSubtract = ImagBuffer[0][0][l] - ImagBuffer[1][0][l];
		stereo_type_detect->subtract_power_y += RealSubtract * RealSubtract + ImagSubtract * ImagSubtract;
		}
		#endif

		/* Compute protos (and their power) for direct sound rendering */

		/* W prototype */
		if ( stereo_type_detect->interpolator > 0 )
		{
		#ifndef FIX_FLOAT_1578_OMASA_REND_SPIKES
		if ( l < ( dipole_freq_range[1] - 1 ) \|\| l >= MASA_SUM_PROTO_START_BIN )
		{
		Real_aux = interpolatorSpaced * 0.5f * Real_aux + interpolatorDmx * Real_aux;
		@@ -1467,6 +1532,59 @@ void protoSignalComputation2(
		p_proto_buffer[2 * l] = interpolatorSpaced * RealBuffer[0][0][l] + interpolatorDmx * Real_aux;
		p_proto_buffer[2 * l + 1] = interpolatorSpaced * ImagBuffer[0][0][l] + interpolatorDmx * Imag_aux;
		}
		#else
		if ( ( stereo_type_detect->type_change_direction == MASA_STEREO_SPACED_MICS && stereo_type_detect->current_stereo_type == MASA_STEREO_DOWNMIX ) \|\|
		( stereo_type_detect->type_change_direction == MASA_STEREO_DOWNMIX && stereo_type_detect->current_stereo_type == MASA_STEREO_SPACED_MICS ) )
		{
		if ( l < ( dipole_freq_range[1] - 1 ) \|\| l >= MASA_SUM_PROTO_START_BIN )
		{
		Real_aux = interpolatorSpaced * 0.5f * Real_aux + interpolatorDmx * Real_aux;
		Imag_aux = interpolatorSpaced * 0.5f * Imag_aux + interpolatorDmx * Imag_aux;
		proto_power_smooth[l] += Real_aux * Real_aux + Imag_aux * Imag_aux;
		p_proto_buffer[2 * l] = Real_aux;
		p_proto_buffer[2 * l + 1] = Imag_aux;
		}
		else
		{
		tempSpaced = RealBuffer[0][0][l] * RealBuffer[0][0][l] + ImagBuffer[0][0][l] * ImagBuffer[0][0][l];
		tempDmx = Real_aux * Real_aux + Imag_aux * Imag_aux;
		proto_power_smooth[l] += interpolatorSpaced * tempSpaced + interpolatorDmx * tempDmx;
		p_proto_buffer[2 * l] = interpolatorSpaced * RealBuffer[0][0][l] + interpolatorDmx * Real_aux;
		p_proto_buffer[2 * l + 1] = interpolatorSpaced * ImagBuffer[0][0][l] + interpolatorDmx * Imag_aux;
		}
		}
		else if ( ( stereo_type_detect->type_change_direction == MASA_STEREO_SPACED_MICS && stereo_type_detect->current_stereo_type == MASA_DUAL_MONO ) \|\|
		( stereo_type_detect->type_change_direction == MASA_DUAL_MONO && stereo_type_detect->current_stereo_type == MASA_STEREO_SPACED_MICS ) )
		{
		if ( l < ( dipole_freq_range[1] - 1 ) \|\| l >= MASA_SUM_PROTO_START_BIN )
		{
		Real_aux *= 0.5f;
		Imag_aux *= 0.5f;
		proto_power_smooth[l] += Real_aux * Real_aux + Imag_aux * Imag_aux;
		p_proto_buffer[2 * l] = Real_aux;
		p_proto_buffer[2 * l + 1] = Imag_aux;
		}
		else
		{
		tempSpaced = RealBuffer[0][0][l] * RealBuffer[0][0][l] + ImagBuffer[0][0][l] * ImagBuffer[0][0][l];
		Real_aux *= 0.5f;
		Imag_aux *= 0.5f;
		tempDualMono = Real_aux * Real_aux + Imag_aux * Imag_aux;
		proto_power_smooth[l] += interpolatorSpaced * tempSpaced + interpolatorDualMono * tempDualMono;
		p_proto_buffer[2 * l] = interpolatorSpaced * RealBuffer[0][0][l] + interpolatorDualMono * Real_aux;
		p_proto_buffer[2 * l + 1] = interpolatorSpaced * ImagBuffer[0][0][l] + interpolatorDualMono * Imag_aux;
		}
		}
		else /* MASA_STEREO_DOWNMIX <-> MASA_DUAL_MONO */
		{
		/* Both use same proto but dual mono has 0.5 scaling */
		Real_aux = interpolatorDualMono * 0.5f * Real_aux + interpolatorDmx * Real_aux;
		Imag_aux = interpolatorDualMono * 0.5f * Imag_aux + interpolatorDmx * Imag_aux;
		proto_power_smooth[l] += Real_aux * Real_aux + Imag_aux * Imag_aux;
		p_proto_buffer[2 * l] = Real_aux;
		p_proto_buffer[2 * l + 1] = Imag_aux;
		}
		#endif
		}
		else if ( stereo_type_detect->masa_stereo_type == MASA_STEREO_SPACED_MICS )
		{
		@@ -1485,7 +1603,19 @@ void protoSignalComputation2(
		p_proto_buffer[2 * l + 1] = ImagBuffer[0][0][l];
		}
		}
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		else if ( stereo_type_detect->masa_stereo_type == MASA_DUAL_MONO )
		{
		Real_aux *= 0.5f;
		Imag_aux *= 0.5f;
		proto_power_smooth[l] += Real_aux * Real_aux + Imag_aux * Imag_aux;
		p_proto_buffer[2 * l] = Real_aux;
		p_proto_buffer[2 * l + 1] = Imag_aux;
		}
		else /* MASA_STEREO_DOWNMIX */
		#else
		else
		#endif
		{
		proto_power_smooth[l] += Real_aux * Real_aux + Imag_aux * Imag_aux;
		p_proto_buffer[2 * l] = Real_aux;
		@@ -1495,6 +1625,7 @@ void protoSignalComputation2(
		/* Y prototype */
		if ( stereo_type_detect->interpolator > 0 )
		{
		#ifndef FIX_FLOAT_1578_OMASA_REND_SPIKES
		if ( l < ( dipole_freq_range[0] ) )
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = interpolatorSpaced * p_proto_buffer[2 * l] + interpolatorDmx * ( RealBuffer[0][0][l] - RealBuffer[1][0][l] );
		@@ -1510,6 +1641,51 @@ void protoSignalComputation2(
		p_proto_buffer[2 * num_freq_bands + 2 * l] = interpolatorSpaced * p_proto_buffer[2 * l] + interpolatorDmx * ( RealBuffer[0][0][l] - RealBuffer[1][0][l] );
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = interpolatorSpaced * p_proto_buffer[2 * l + 1] + interpolatorDmx * ( ImagBuffer[0][0][l] - ImagBuffer[1][0][l] );
		}
		#else
		if ( ( stereo_type_detect->type_change_direction == MASA_STEREO_SPACED_MICS && stereo_type_detect->current_stereo_type == MASA_STEREO_DOWNMIX ) \|\|
		( stereo_type_detect->type_change_direction == MASA_STEREO_DOWNMIX && stereo_type_detect->current_stereo_type == MASA_STEREO_SPACED_MICS ) )
		{
		if ( l < ( dipole_freq_range[0] ) )
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = interpolatorSpaced * p_proto_buffer[2 * l] + interpolatorDmx * ( RealBuffer[0][0][l] - RealBuffer[1][0][l] );
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = interpolatorSpaced * p_proto_buffer[2 * l + 1] + interpolatorDmx * ( ImagBuffer[0][0][l] - ImagBuffer[1][0][l] );
		}
		else if ( l < ( dipole_freq_range[1] ) )
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = interpolatorSpaced * ( ImagBuffer[0][0][l] - ImagBuffer[1][0][l] ) + interpolatorDmx * ( RealBuffer[0][0][l] - RealBuffer[1][0][l] );
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = interpolatorSpaced * ( -( RealBuffer[0][0][l] - RealBuffer[1][0][l] ) ) + interpolatorDmx * ( ImagBuffer[0][0][l] - ImagBuffer[1][0][l] );
		}
		else
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = interpolatorSpaced * p_proto_buffer[2 * l] + interpolatorDmx * ( RealBuffer[0][0][l] - RealBuffer[1][0][l] );
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = interpolatorSpaced * p_proto_buffer[2 * l + 1] + interpolatorDmx * ( ImagBuffer[0][0][l] - ImagBuffer[1][0][l] );
		}
		}
		else if ( ( stereo_type_detect->type_change_direction == MASA_STEREO_SPACED_MICS && stereo_type_detect->current_stereo_type == MASA_DUAL_MONO ) \|\|
		( stereo_type_detect->type_change_direction == MASA_DUAL_MONO && stereo_type_detect->current_stereo_type == MASA_STEREO_SPACED_MICS ) )
		{
		if ( l < ( dipole_freq_range[0] ) )
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = p_proto_buffer[2 * l];
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = p_proto_buffer[2 * l + 1];
		}
		else if ( l < ( dipole_freq_range[1] ) )
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = interpolatorSpaced * ( ImagBuffer[0][0][l] - ImagBuffer[1][0][l] ) + interpolatorDualMono * p_proto_buffer[2 * l];
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = interpolatorSpaced * ( -( RealBuffer[0][0][l] - RealBuffer[1][0][l] ) ) + interpolatorDualMono * p_proto_buffer[2 * l + 1];
		}
		else
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = p_proto_buffer[2 * l];
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = p_proto_buffer[2 * l + 1];
		}
		}
		else /* MASA_STEREO_DOWNMIX <-> MASA_DUAL_MONO */
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = interpolatorDualMono * p_proto_buffer[2 * l] + interpolatorDmx * ( RealBuffer[0][0][l] - RealBuffer[1][0][l] );
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = interpolatorDualMono * p_proto_buffer[2 * l + 1] + interpolatorDmx * ( ImagBuffer[0][0][l] - ImagBuffer[1][0][l] );
		}
		#endif
		proto_power_smooth[l + num_freq_bands] += p_proto_buffer[2 * num_freq_bands + 2 * l] * p_proto_buffer[2 * num_freq_bands + 2 * l] + p_proto_buffer[2 * num_freq_bands + 2 * l + 1] * p_proto_buffer[2 * num_freq_bands + 2 * l + 1];
		}
		else if ( stereo_type_detect->masa_stereo_type == MASA_STEREO_SPACED_MICS )
		@@ -1533,7 +1709,18 @@ void protoSignalComputation2(
		proto_power_smooth[l + num_freq_bands] = proto_power_smooth[l];
		}
		}
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		else if ( stereo_type_detect->masa_stereo_type == MASA_DUAL_MONO )
		{
		/* proto = W */
		p_proto_buffer[2 * num_freq_bands + 2 * l] = p_proto_buffer[2 * l];
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = p_proto_buffer[2 * l + 1];
		proto_power_smooth[l + num_freq_bands] = proto_power_smooth[l];
		}
		else /* MASA_STEREO_DOWNMIX */
		#else
		else
		#endif
		{
		p_proto_buffer[2 * num_freq_bands + 2 * l] = RealBuffer[0][0][l] - RealBuffer[1][0][l];
		p_proto_buffer[2 * num_freq_bands + 2 * l + 1] = ImagBuffer[0][0][l] - ImagBuffer[1][0][l];
		@@ -1821,7 +2008,12 @@ void computeDirectionAngles(
		------------------------------------------------------------------------/

		void ivas_masa_init_stereotype_detection(
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		MASA_STEREO_TYPE_DETECT *stereo_type_detect,
		int32_t ivas_total_brate )
		#else
		MASA_STEREO_TYPE_DETECT *stereo_type_detect )
		#endif
		{
		stereo_type_detect->masa_stereo_type = MASA_STEREO_DOWNMIX;
		stereo_type_detect->current_stereo_type = MASA_STEREO_DOWNMIX;
		@@ -1853,6 +2045,17 @@ void ivas_masa_init_stereotype_detection(
		stereo_type_detect->min_sum_total_ratio_db = 0.0f;
		stereo_type_detect->subtract_target_ratio_db = 0.0f;

		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		stereo_type_detect->inst_diff_bb_power = 0.0f;
		stereo_type_detect->inst_total_bb_power = 0.0f;

		stereo_type_detect->max_band_diff_ene = IVAS_CLDFB_NO_CHANNELS_MAX;
		if ( ivas_total_brate < IVAS_48k )
		{
		stereo_type_detect->max_band_diff_ene = 18;
		}

		#endif
		return;
		}

		@@ -1879,7 +2082,23 @@ void ivas_masa_stereotype_detection(
		float min_sum_temp;
		float lr_total_bb_temp;
		float lr_total_hi_temp;
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		int16_t change_to_dual_mono_selection;
		float diffPerSum;

		/* Determine if broadband energy and ratio between difference broadband energy and broadband energy indicate
		* that the signal type is dual mono */
		change_to_dual_mono_selection = 0;
		if ( stereo_type_detect->inst_total_bb_power > 1.0f )
		{
		diffPerSum = stereo_type_detect->inst_diff_bb_power / stereo_type_detect->inst_total_bb_power;
		if ( diffPerSum < ( stereo_type_detect->current_stereo_type == MASA_DUAL_MONO ? MASA_DUAL_MONO_TAU1 : MASA_DUAL_MONO_TAU2 ) )
		{
		change_to_dual_mono_selection = 1;
		}
		}

		#endif
		/* Determine if the determined features match the spaced mic type */
		change_to_spaced_selection = 0;
		if ( subtract_target_ratio_db < -3.0f )
		@@ -1933,7 +2152,15 @@ void ivas_masa_stereotype_detection(
		}
		else
		{
		#ifdef FIX_FLOAT_1578_OMASA_REND_SPIKES
		if ( change_to_dual_mono_selection == 1 )
		{
		stereo_type_detect->masa_stereo_type = MASA_DUAL_MONO;
		}
		else if ( change_to_spaced_selection == 1 )
		#else
		if ( change_to_spaced_selection == 1 )
		#endif
		{
		stereo_type_detect->masa_stereo_type = MASA_STEREO_SPACED_MICS;
		}