Merge branch '331-hbr-low-frequency-stuttering-fix' into 'main' (f7278d9d) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_cov_smooth.c

+115 −15

Original line number	Diff line number	Diff line
		@@ -39,7 +39,9 @@
		#include "wmc_auto.h"
		#include "prot.h"


		#ifdef SMOOTH_WITH_TRANS_DET
		#define BAND_SMOOTH_REST_START_IDX ( 2 )
		#endif
		/-----------------------------------------------------------------------------------------
		* Function ivas_set_up_cov_smoothing()
		*
		@@ -50,13 +52,23 @@ static void ivas_set_up_cov_smoothing(
		ivas_cov_smooth_state_t *hCovState,
		ivas_filterbank_t *pFb,
		const float max_update_rate,
		const int16_t min_pool_size )
		const int16_t min_pool_size
		#ifdef COV_SMOOTH_TUNING
		,
		const int16_t nchan_inp /* i : number of input channels */
		#endif
		)
		{
		int16_t j, k;

		#ifdef COV_SMOOTH_TUNING
		if ( nchan_inp <= FOA_CHANNELS )
		{
		#endif
		for ( j = 0; j < pFb->filterbank_num_bands; j++ )
		{
		float update_factor = 0.0f;
		float update_factor;
		update_factor = 0.0f;

		for ( k = 0; k < pFb->fb_bin_to_band.pFb_active_bins_per_band[j]; k++ )
		{
		@@ -71,6 +83,31 @@ static void ivas_set_up_cov_smoothing(
		}
		}

		#ifdef COV_SMOOTH_TUNING
		}
		else
		{
		for ( j = 0; j < pFb->filterbank_num_bands; j++ )
		{
		float update_factor;
		float *p_bin_to_band;
		int16_t active_bins;
		update_factor = 0.0f;
		p_bin_to_band = pFb->fb_bin_to_band.pp_short_stride_bin_to_band[j];
		active_bins = pFb->fb_bin_to_band.p_short_stride_num_bins_per_band[j];
		for ( k = 0; k < active_bins; k++ )
		{
		update_factor += p_bin_to_band[k];
		}
		hCovState->pSmoothing_factor[j] = update_factor / min_pool_size;
		hCovState->pSmoothing_factor[j] = ( j + 1 ) 0.75f;
		if ( hCovState->pSmoothing_factor[j] > max_update_rate )
		{
		hCovState->pSmoothing_factor[j] = max_update_rate;
		}
		}
		}
		#endif
		hCovState->prior_bank_idx = -1;

		return;
		@@ -115,7 +152,12 @@ ivas_error ivas_spar_covar_smooth_enc_open(
		}
		}

		ivas_set_up_cov_smoothing( hCovState, pFb, cov_smooth_cfg->max_update_rate, cov_smooth_cfg->min_pool_size );
		ivas_set_up_cov_smoothing( hCovState, pFb, cov_smooth_cfg->max_update_rate, cov_smooth_cfg->min_pool_size
		#ifdef COV_SMOOTH_TUNING
		,
		nchan_inp
		#endif
		);

		*hCovState_out = hCovState;

		@@ -176,14 +218,66 @@ static void ivas_compute_smooth_cov(
		const int16_t start_band,
		const int16_t end_band,
		const int16_t num_ch,
		const int16_t transient_det )
		#ifdef SMOOTH_WITH_TRANS_DET
		const int16_t transient_det[2]
		#else
		const int16_t transient_det
		#endif
		)
		{
		int16_t i, j, k;
		int16_t prev_idx = hCovState->prior_bank_idx;
		float factor = 0;
		#ifdef SMOOTH_WITH_TRANS_DET
		int16_t sm_b;
		int16_t non_sm_b_idx;
		sm_b = BAND_SMOOTH_REST_START_IDX;
		#endif

		assert( end_band <= pFb->filterbank_num_bands );

		#ifdef SMOOTH_WITH_TRANS_DET
		if ( prev_idx == -1 \|\| transient_det[1] == 1 )
		{
		for ( i = 0; i < num_ch; i++ )
		{
		for ( k = start_band; k < end_band; k++ )
		{
		pCov_buf[i][i][k] += ( hCovState->pSmoothing_factor[k] * fac );
		}
		}
		}
		else if ( transient_det[0] == 1 )
		{
		non_sm_b_idx = min( sm_b, end_band );
		for ( i = 0; i < num_ch; i++ )
		{
		for ( j = 0; j < num_ch; j++ )
		{
		if ( i == j )
		{
		factor = fac;
		}
		else
		{
		factor = 0.0f;
		}

		for ( k = start_band; k < non_sm_b_idx; k++ )
		{
		pCov_buf[i][j][k] = pPrior_cov_buf[i][j][k] + ( hCovState->pSmoothing_factor[k] * ( pCov_buf[i][j][k] - pPrior_cov_buf[i][j][k] + factor ) );
		}
		}
		}
		for ( i = 0; i < num_ch; i++ )
		{
		for ( k = non_sm_b_idx; k < end_band; k++ )
		{
		pCov_buf[i][i][k] += ( hCovState->pSmoothing_factor[k] * fac );
		}
		}
		}
		#else
		if ( prev_idx == -1 \|\| transient_det == 1 )
		{
		for ( i = 0; i < num_ch; i++ )
		@@ -194,6 +288,7 @@ static void ivas_compute_smooth_cov(
		}
		}
		}
		#endif
		else if ( prev_idx == 0 )
		{
		for ( i = 0; i < num_ch; i++ )
		@@ -234,7 +329,12 @@ void ivas_cov_smooth_process(
		const int16_t start_band,
		const int16_t end_band,
		const int16_t num_ch,
		const int16_t transient_det )
		#ifdef SMOOTH_WITH_TRANS_DET
		const int16_t transient_det[2]
		#else
		const int16_t transient_det
		#endif
		)
		{
		int16_t i, j;
		int16_t num_bands = end_band - start_band;

lib_com/ivas_prot.h

+17 −0

Original line number	Diff line number	Diff line
		@@ -4189,7 +4189,11 @@ void ivas_enc_cov_handler_process(
		const int16_t end_band,
		const int16_t nchan_inp,
		const int16_t dtx_vad,
		#ifdef SMOOTH_WITH_TRANS_DET
		const int16_t transient_det[2]
		#else
		const int16_t transient_det
		#endif
		);

		ivas_error ivas_spar_covar_smooth_enc_open(
		@@ -4211,7 +4215,11 @@ void ivas_cov_smooth_process(
		const int16_t start_band,
		const int16_t end_band,
		const int16_t num_ch,
		#ifdef SMOOTH_WITH_TRANS_DET
		const int16_t transient_det[2]
		#else
		const int16_t transient_det
		#endif
		);

		/* Transient detector module */
		@@ -4224,11 +4232,20 @@ void ivas_spar_transient_det_close(
		ivas_trans_det_state_t *hTranDet / i/o: SPAR TD handle */
		);

		#ifdef SMOOTH_WITH_TRANS_DET
		void ivas_transient_det_process(
		ivas_trans_det_state_t hTranDet, / i/o: SPAR TD handle */
		float pIn_pcm, / i : input audio channels */
		const int16_t frame_len, /* i : frame length in samples */
		int16_t transient_det[2] /* o: transient det outputs */
		);
		#else
		int16_t ivas_transient_det_process(
		ivas_trans_det_state_t hTranDet, / i/o: SPAR TD handle */
		float pIn_pcm, / i : input audio channels */
		const int16_t frame_len /* i : frame length in samples */
		);
		#endif

		void ivas_td_decorr_get_ducking_gains(
		ivas_trans_det_state_t hTranDet, / i/o: SPAR TD handle */

lib_com/ivas_transient_det.c

+44 −2

Original line number	Diff line number	Diff line
		@@ -211,16 +211,31 @@ void ivas_spar_transient_det_close(
		*
		* Transient detection process call
		-----------------------------------------------------------------------------------------/

		#ifdef SMOOTH_WITH_TRANS_DET
		void ivas_transient_det_process(
		ivas_trans_det_state_t hTranDet, / i/o: SPAR TD handle */
		float pIn_pcm, / i : input audio channels */
		const int16_t frame_len, /* i : frame length in samples */
		int16_t transient_det[2] /* o: transient det outputs */
		)
		#else
		int16_t ivas_transient_det_process(
		ivas_trans_det_state_t hTranDet, / i/o: SPAR TD handle */
		float pIn_pcm, / i : input audio channels */
		const int16_t frame_len /* i : frame length in samples */
		)
		#endif
		{
		float mem = hTranDet->in_duck_gain;
		#ifdef SMOOTH_WITH_TRANS_DET
		float in_duck_gain[L_FRAME48k];
		int16_t num_sf, sf, sf_samp, idx;
		#else
		int16_t trans;
		#endif
		float mem = hTranDet->in_duck_gain;


		#ifndef SMOOTH_WITH_TRANS_DET
		ivas_td_decorr_get_ducking_gains( hTranDet, pIn_pcm, NULL, NULL, frame_len, IVAS_TDET_ONLY );

		if ( mem - hTranDet->in_duck_gain > IVAS_TDET_PARM_TRANS_THR )
		@@ -233,6 +248,30 @@ int16_t ivas_transient_det_process(
		}

		return trans;
		#else
		ivas_td_decorr_get_ducking_gains( hTranDet, pIn_pcm, in_duck_gain, NULL, frame_len, IVAS_TDET_ONLY );

		transient_det[0] = 0;
		transient_det[1] = 0;
		if ( mem - hTranDet->in_duck_gain > IVAS_TDET_PARM_TRANS_THR )
		{
		transient_det[0] = 1;
		}

		num_sf = 16;
		sf_samp = frame_len / num_sf;
		for ( sf = 1; sf <= num_sf; sf++ )
		{
		idx = ( sf_samp * sf ) - 1;
		if ( ( mem - in_duck_gain[idx] ) > ( IVAS_TDET_PARM_TRANS_THR * 1.1f ) )
		{
		transient_det[1] = 1;
		}
		mem = in_duck_gain[idx];
		}
		// dbgwrite( &transient_det[1], sizeof( int16_t ), 1, 1, "trans_det.raw" );
		return;
		#endif
		}


		@@ -306,6 +345,9 @@ void ivas_td_decorr_get_ducking_gains(
		for ( i = 0; i < frame_len; i++ )
		{
		in_duck_gain = ivas_calc_duck_gain( in_duck_gain, in_duck_coeff, e_slow[i], e_fast[i], duck_mult_fac );
		#ifdef SMOOTH_WITH_TRANS_DET
		pIn_duck_gains[i] = in_duck_gain;
		#endif
		}
		hTranDet->in_duck_gain = in_duck_gain;
		}

lib_com/options.h

+8 −0

Original line number	Diff line number	Diff line
		@@ -167,6 +167,14 @@

		#define FIX_107_5MS_SUBFRAME_RENDERING /* Issue 107: use 5ms subframes everywhere in parametric binauralizer */

		#define FIX_331_SBA_HBR_HOA_FIXES /* DLB: issue 331 - fix addressing low frequency stuttering with HOA inputs at high bitrates */
		#ifdef FIX_331_SBA_HBR_HOA_FIXES
		#define COV_SMOOTH_TUNING
		#define SBA_HPF_TUNING_ENC
		/#define SBA_HPF_TUNING_DEC/
		#define SMOOTH_WITH_TRANS_DET
		#endif

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

lib_dec/ivas_dec.c

+13 −0

Original line number	Diff line number	Diff line
		@@ -327,7 +327,20 @@ ivas_error ivas_dec(

		/* HP filtering */
		#ifndef DEBUG_SPAR_BYPASS_EVS_CODEC
		#ifdef SBA_HPF_TUNING_DEC
		int16_t nchan_hp20;
		if ( st_ivas->ivas_format == MASA_FORMAT )
		{
		nchan_hp20 = nchan_remapped;
		}
		else
		{
		nchan_hp20 = getNumChanSynthesis( st_ivas );
		}
		for ( n = 0; n < nchan_hp20; n++ )
		#else
		for ( n = 0; n < nchan_remapped; n++ )
		#endif
		{
		hp20( output[n], output_frame, st_ivas->mem_hp20_out[n], output_Fs );
		}