Implementation of contribution 28 - adaptive transport audio signals for head tracked param-bin (8eee1939) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/options.h

+3 −1

Original line number	Diff line number	Diff line
		@@ -158,6 +158,8 @@
		#define SMOOTH_WITH_TRANS_DET
		#endif

		#define NOKIA_ADAPTIVE_BINAURAL_PROTOS

		/* ################## End DEVELOPMENT switches ######################### */
		/* clang-format on */
		#endif

lib_rend/ivas_dirac_dec_binaural_functions.c

+95 −0

Original line number	Diff line number	Diff line
		@@ -54,6 +54,14 @@
		#define IVAS_TDET_DUCK_MULT_FAC_PARA_BIN ( 2.0f )
		#define IVAS_TDET_DUCK_MULT_FAC_PARA_BIN_LOW_BR ( 3.0f )

		#ifdef NOKIA_ADAPTIVE_BINAURAL_PROTOS
		#define ADAPT_HTPROTO_IIR_FAC 0.95f
		#define ADAPT_HTPROTO_ILD_LIM_DB0 1.0f
		#define ADAPT_HTPROTO_ILD_LIM_DB1 4.0f
		#define ADAPT_HTPROTO_ROT_LIM_0 0.4f
		#define ADAPT_HTPROTO_ROT_LIM_1 0.8f
		#endif

		/*-------------------------------------------------------------------------
		* Local function prototypes
		------------------------------------------------------------------------/
		@@ -68,6 +76,10 @@ static void ivas_dirac_dec_binaural_determine_processing_matrices( Decoder_Struc

		static void ivas_dirac_dec_binaural_process_output( Decoder_Struct *st_ivas, float output_f[][L_FRAME48k], float inRe[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX], float inIm[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX], const int16_t max_band_decorr, const uint8_t numInputChannels, const uint8_t firstSlot, const uint8_t slotEnd );

		#ifdef NOKIA_ADAPTIVE_BINAURAL_PROTOS
		static void adaptTransportSignalsHeadtracked( HEAD_TRACK_DATA_HANDLE hHeadTrackData, float inIm[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX], float inRe[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX], const uint8_t firstSlot, const uint8_t slotEnd, const uint8_t nBins, float Rmat[3][3] );
		#endif

		static void ivas_dirac_dec_binaural_check_and_switch_transports_headtracked( HEAD_TRACK_DATA_HANDLE hHeadTrackData, float inIm[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX], float inRe[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX], const uint8_t firstSlot, const uint8_t slotEnd, const uint8_t nBins, float Rmat[3][3] );

		static void formulate2x2MixingMatrix( float Ein1, float Ein2, float CinRe, float CinIm, float Eout1, float Eout2, float CoutRe, float CoutIm, float Q[BINAURAL_CHANNELS][BINAURAL_CHANNELS], float Mre[BINAURAL_CHANNELS][BINAURAL_CHANNELS], float Mim[BINAURAL_CHANNELS][BINAURAL_CHANNELS], const float regularizationFactor );
		@@ -562,6 +574,10 @@ static void ivas_dirac_dec_binaural_internal(

		if ( nchan_transport == 2 )
		{
		#ifdef NOKIA_ADAPTIVE_BINAURAL_PROTOS
		adaptTransportSignalsHeadtracked( st_ivas->hHeadTrackData, Cldfb_RealBuffer_in, Cldfb_ImagBuffer_in, firstSlot, slotEnd, nBins, Rmat );
		#endif

		ivas_dirac_dec_binaural_check_and_switch_transports_headtracked( st_ivas->hHeadTrackData, Cldfb_ImagBuffer_in, Cldfb_RealBuffer_in, firstSlot, slotEnd, nBins, Rmat );
		}
		}
		@@ -1332,6 +1348,85 @@ static void ivas_dirac_dec_binaural_process_output(
		}


		#ifdef NOKIA_ADAPTIVE_BINAURAL_PROTOS
		static void adaptTransportSignalsHeadtracked(
		HEAD_TRACK_DATA_HANDLE hHeadTrackData,
		float inIm[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX],
		float inRe[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX],
		const uint8_t firstSlot,
		const uint8_t slotEnd,
		const uint8_t nBins,
		float Rmat[3][3] )
		{
		int16_t slot, ch, bin, louderCh;
		float re[2], im[2], ILD, mono_factor_ILD, mono_factor_rotation, mono_factor, y_val;
		float proc_re[2], proc_im[2], sum_re, sum_im, ene_proc, ene_target, mf;

		/* Determine head-orientation-based mono factor.
		Rmat[1][1] entry informs how close the ears are aligned according to transport signals. */
		y_val = 1.0f - fabsf( Rmat[1][1] );
		mono_factor_rotation = ( y_val - ADAPT_HTPROTO_ROT_LIM_0 ) / ( ADAPT_HTPROTO_ROT_LIM_1 - ADAPT_HTPROTO_ROT_LIM_0 );
		mono_factor_rotation = fmaxf( 0.0f, fminf( 1.0f, mono_factor_rotation ) );

		/* Adapt transport signals in frequency bands */
		for ( slot = firstSlot; slot < slotEnd; slot++ )
		{
		float eqVal[60];
		for ( bin = 0; bin < nBins; bin++ )
		{
		/* Determine channel energies */
		for ( ch = 0; ch < 2; ch++ )
		{
		re[ch] = inRe[ch][slot][bin];
		im[ch] = inIm[ch][slot][bin];

		hHeadTrackData->chEneIIR[ch][bin] *= ADAPT_HTPROTO_IIR_FAC;
		hHeadTrackData->chEneIIR[ch][bin] += ( 1.0f - ADAPT_HTPROTO_IIR_FAC ) * ( ( re[ch] * re[ch] ) + ( im[ch] * im[ch] ) );
		}

		/* Determine ILD */
		ILD = fabsf( 10.0f * log10f( fmaxf( 1e-12f, hHeadTrackData->chEneIIR[0][bin] ) / fmaxf( 1e-12f, hHeadTrackData->chEneIIR[1][bin] ) ) );
		louderCh = ( hHeadTrackData->chEneIIR[1][bin] > hHeadTrackData->chEneIIR[0][bin] );

		/* Determine ILD-based mono factor */
		mono_factor_ILD = ( ILD - ADAPT_HTPROTO_ILD_LIM_DB0 ) / ( ADAPT_HTPROTO_ILD_LIM_DB1 - ADAPT_HTPROTO_ILD_LIM_DB0 );
		mono_factor_ILD = fmaxf( 0.0f, fminf( 1.0f, mono_factor_ILD ) );

		/* Combine mono factors */
		mono_factor = mono_factor_ILD * mono_factor_rotation;

		/* Mix original audio and sum signal according to determined mono factor */
		sum_re = re[0] + re[1];
		sum_im = im[0] + im[1];
		for ( ch = 0; ch < 2; ch++ )
		{
		mf = ( ch == louderCh ) ? 0.0f : mono_factor;

		proc_re[ch] = mf * sum_re + ( 1.0f - mf ) * re[ch];
		proc_im[ch] = mf * sum_im + ( 1.0f - mf ) * im[ch];

		hHeadTrackData->procChEneIIR[ch][bin] *= ADAPT_HTPROTO_IIR_FAC;
		hHeadTrackData->procChEneIIR[ch][bin] += ( 1.0f - ADAPT_HTPROTO_IIR_FAC ) * ( ( proc_re[ch] * proc_re[ch] ) + ( proc_im[ch] * proc_im[ch] ) );
		}

		/* Equalize */
		ene_target = hHeadTrackData->chEneIIR[0][bin] + hHeadTrackData->chEneIIR[1][bin];
		ene_proc = hHeadTrackData->procChEneIIR[0][bin] + hHeadTrackData->procChEneIIR[1][bin];
		eqVal[bin] = fminf( 4.0f, sqrtf( ene_target / fmaxf( 1e-12f, ene_proc ) ) );

		for ( ch = 0; ch < 2; ch++ )
		{
		inRe[ch][slot][bin] = proc_re[ch] * eqVal[bin];
		inIm[ch][slot][bin] = proc_im[ch] * eqVal[bin];
		}
		}
		}

		return;
		}
		#endif


		static void ivas_dirac_dec_binaural_check_and_switch_transports_headtracked(
		HEAD_TRACK_DATA_HANDLE hHeadTrackData,
		float inIm[][CLDFB_NO_COL_MAX][CLDFB_NO_CHANNELS_MAX],

lib_rend/ivas_hrtf.c

+7 −0

Original line number	Diff line number	Diff line
		@@ -235,5 +235,12 @@ ivas_error ivas_headTrack_open(
		( *hHeadTrackData )->Rmat_prev[i][i] = 1.0f;
		}

		#ifdef NOKIA_ADAPTIVE_BINAURAL_PROTOS
		set_zero( ( *hHeadTrackData )->chEneIIR[0], CLDFB_NO_CHANNELS_MAX );
		set_zero( ( *hHeadTrackData )->chEneIIR[1], CLDFB_NO_CHANNELS_MAX );
		set_zero( ( *hHeadTrackData )->procChEneIIR[0], CLDFB_NO_CHANNELS_MAX );
		set_zero( ( *hHeadTrackData )->procChEneIIR[1], CLDFB_NO_CHANNELS_MAX );
		#endif

		return IVAS_ERR_OK;
		}

lib_rend/ivas_stat_rend.h

+5 −0

Original line number	Diff line number	Diff line
		@@ -245,6 +245,11 @@ typedef struct ivas_binaural_head_track_struct
		uint8_t lrSwitchedCurrent;
		float lrSwitchInterpVal;

		#ifdef NOKIA_ADAPTIVE_BINAURAL_PROTOS
		float chEneIIR[2][CLDFB_NO_CHANNELS_MAX];
		float procChEneIIR[2][CLDFB_NO_CHANNELS_MAX];
		#endif

		int16_t shd_rot_max_order;

		} HEAD_TRACK_DATA, *HEAD_TRACK_DATA_HANDLE;

lib_dec/ivas_stat_dec.h

+2 −2

File changed.

Contains only whitespace changes.