Issue 549: Downmix energy compensation for ParamISM to raise output levels. (ad1a2a99) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_ism_com.c

+4 −0

Original line number	Diff line number	Diff line
		@@ -483,6 +483,10 @@ void ivas_param_ism_config(
		hParamIsm->last_el_sgn[i] = 1;
		}

		#ifdef FIX_549_DMX_GAIN
		hParamIsm->last_dmx_gain = 1.0f;
		set_f( hParamIsm->last_cardioid_left, 1.0f, MAX_NUM_OBJECTS );
		#endif
		return;
		}

lib_com/ivas_stat_com.h

+5 −0

Original line number	Diff line number	Diff line
		@@ -169,6 +169,11 @@ typedef struct ivas_param_ism_data_structure
		int16_t noisy_speech_buffer[PARAM_ISM_HYS_BUF_SIZE];
		int16_t flag_equal_energy;

		#ifdef FIX_549_DMX_GAIN
		float last_dmx_gain;
		float last_cardioid_left[MAX_NUM_OBJECTS];
		#endif

		} PARAM_ISM_CONFIG_DATA, *PARAM_ISM_CONFIG_HANDLE;

lib_com/options.h

+2 −0

Original line number	Diff line number	Diff line
		@@ -185,6 +185,8 @@

		#define FIX_528_ISM_MD_FILE_TOO_SHORT /* VA: issue 528: ISM Metadata file too short */

		#define FIX_549_DMX_GAIN /* FhG: issue 549: ParamISM output too quiet */


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

lib_dec/ivas_ism_param_dec.c

+94 −0

Original line number	Diff line number	Diff line
		@@ -729,6 +729,11 @@ void ivas_param_ism_dec(
		int16_t ch, nchan_transport, nchan_out, nchan_out_woLFE, i;
		int16_t subframe_idx, slot_idx, index_slot, bin_idx;
		int32_t ivas_total_brate;
		#ifdef FIX_549_DMX_GAIN
		int16_t output_frame;
		float gain, ene_tc, ene_sum, grad;
		float *last_gain;
		#endif
		float ref_power[CLDFB_NO_CHANNELS_MAX];
		float cx_diag[CLDFB_NO_CHANNELS_MAX][PARAM_ISM_MAX_DMX];
		/* CLDFB Input Buffers */
		@@ -752,6 +757,12 @@ void ivas_param_ism_dec(
		/* Initialization */
		hDirAC = st_ivas->hDirAC;
		assert( hDirAC );
		#ifdef FIX_549_DMX_GAIN
		ene_tc = 0.0f;
		ene_sum = 0.0f;
		last_gain = &( st_ivas->hDirAC->hParamIsm->last_dmx_gain );
		output_frame = (int16_t) ( st_ivas->hDecoderConfig->output_Fs / FRAMES_PER_SEC );
		#endif

		nchan_transport = st_ivas->nchan_transport;
		if ( st_ivas->hDecoderConfig->output_config == AUDIO_CONFIG_EXTERNAL )
		@@ -835,6 +846,42 @@ void ivas_param_ism_dec(
		}
		}

		#ifdef FIX_549_DMX_GAIN
		/* Energy Compensation */
		for ( i = 0; i < output_frame; i++ )
		{
		ene_tc += output_f[0][i] * output_f[0][i] + output_f[1][i] * output_f[1][i]; // LL + RR
		ene_sum += ( output_f[0][i] + output_f[1][i] ) * ( output_f[0][i] + output_f[1][i] ); // (L+R)*(L+R)
		}
		gain = sqrtf( ene_tc / ( ene_sum + EPSILON ) );
		if ( st_ivas->hSCE[0]->hCoreCoder[0]->ini_frame > 1 )
		{
		/* Smoothing */
		gain = 0.75f * gain + 0.25f * *( last_gain );
		/* 10ms ramp */
		grad = ( gain - ( last_gain ) ) 2.0f / output_frame; /* slope between two consecutive gains, 480 samples length */
		for ( i = 0; i < ( output_frame / 2 ); i++ )
		{
		output_f[0][i] = ( ( last_gain ) + i * grad );
		output_f[1][i] = ( ( last_gain ) + i * grad );
		}
		for ( ; i < output_frame; i++ )
		{
		output_f[0][i] *= gain;
		output_f[1][i] *= gain;
		}
		}
		else
		{
		for ( i = 0; i < output_frame; i++ )
		{
		output_f[0][i] *= gain;
		output_f[1][i] *= gain;
		}
		}
		*( last_gain ) = gain;
		#endif

		for ( ch = 0; ch < nchan_transport; ch++ )
		{
		/-----------------------------------------------------------------
		@@ -1055,6 +1102,11 @@ void ivas_param_ism_dec_digest_tc(
		int16_t ch, nchan_transport, nchan_out, nchan_out_woLFE, i;
		int16_t slot_idx, bin_idx;
		int32_t ivas_total_brate;
		#ifdef FIX_549_DMX_GAIN
		int16_t output_frame;
		float gain, ene_tc, ene_sum, grad;
		float *last_gain;
		#endif
		float ref_power[CLDFB_NO_CHANNELS_MAX];
		float cx_diag[CLDFB_NO_CHANNELS_MAX][PARAM_ISM_MAX_DMX];
		/* Direct Response/EFAP Gains */
		@@ -1064,6 +1116,12 @@ void ivas_param_ism_dec_digest_tc(
		/* Initialization */
		hDirAC = st_ivas->hDirAC;
		assert( hDirAC );
		#ifdef FIX_549_DMX_GAIN
		ene_tc = 0.0f;
		ene_sum = 0.0f;
		last_gain = &( st_ivas->hDirAC->hParamIsm->last_dmx_gain );
		output_frame = (int16_t) ( st_ivas->hDecoderConfig->output_Fs / FRAMES_PER_SEC );
		#endif

		nchan_transport = st_ivas->nchan_transport;
		ivas_total_brate = st_ivas->hDecoderConfig->ivas_total_brate;
		@@ -1150,6 +1208,42 @@ void ivas_param_ism_dec_digest_tc(
		}
		}

		#ifdef FIX_549_DMX_GAIN
		/* Energy Compensation */
		for ( i = 0; i < output_frame; i++ )
		{
		ene_tc += transport_channels_f[0][i] * transport_channels_f[0][i] + transport_channels_f[1][i] * transport_channels_f[1][i]; // LL + RR
		ene_sum += ( transport_channels_f[0][i] + transport_channels_f[1][i] ) * (transport_channels_f[0][i] + transport_channels_f[1][i] ); // (L+R)*(L+R)
		}
		gain = sqrtf( ene_tc / ( ene_sum + EPSILON ) );
		if ( st_ivas->hSCE[0]->hCoreCoder[0]->ini_frame > 1 )
		{
		/* Smoothing */
		gain = 0.75f * gain + 0.25f * *( last_gain );
		/* 10ms ramp */
		grad = ( gain - ( last_gain ) ) 2.0f / output_frame; /* slope between two consecutive gains, 480 samples length */
		for ( i = 0; i < ( output_frame / 2 ); i++ )
		{
		transport_channels_f[0][i] = ( ( last_gain ) + i * grad );
		transport_channels_f[1][i] = ( ( last_gain ) + i * grad );
		}
		for ( ; i < output_frame; i++ )
		{
		transport_channels_f[0][i] *= gain;
		transport_channels_f[1][i] *= gain;
		}
		}
		else
		{
		for ( i = 0; i < output_frame; i++ )
		{
		transport_channels_f[0][i] *= gain;
		transport_channels_f[1][i] *= gain;
		}
		}
		*( last_gain ) = gain;
		#endif

		for ( ch = 0; ch < nchan_transport; ch++ )
		{
		/-----------------------------------------------------------------

lib_enc/ivas_ism_param_enc.c

+88 −0

Original line number	Diff line number	Diff line
		@@ -222,6 +222,11 @@ void ivas_param_ism_stereo_dmx(
		float alpha, azi_shift, tmp, tmp_1;
		float cardioid_left[MAX_NUM_OBJECTS], cardioid_right[MAX_NUM_OBJECTS];
		float stereo_dmx[2][L_FRAME48k];
		#ifdef FIX_549_DMX_GAIN
		float dmx_gain, ene_dmx, ene_data, grad;
		float *last_dmx_gain;
		float *last_cardioid_left;
		#endif
		ISM_METADATA_HANDLE hIsmMetaData;

		push_wmops( "ivas_param_ism_st_dmx" );
		@@ -229,6 +234,13 @@ void ivas_param_ism_stereo_dmx(
		/Initialization/
		alpha = 0.5;
		azi_shift = 0;
		#ifdef FIX_549_DMX_GAIN
		dmx_gain = 0;
		ene_dmx = 0;
		ene_data = 0;
		last_dmx_gain = &( st_ivas->hDirAC->hParamIsm->last_dmx_gain );
		last_cardioid_left = &( st_ivas->hDirAC->hParamIsm->last_cardioid_left[0] );
		#endif

		/* Set the stereo dmx to zero */
		set_zero( stereo_dmx[0], L_FRAME48k );
		@@ -243,6 +255,7 @@ void ivas_param_ism_stereo_dmx(
		tmp = hIsmMetaData->azimuth * ( EVS_PI / 180 );
		tmp_1 = ( EVS_PI / 2 ) + azi_shift;
		cardioid_left[i] = alpha + ( 1 - alpha ) * cosf( tmp - tmp_1 );
		#ifndef FIX_549_DMX_GAIN
		cardioid_right[i] = alpha + ( 1 - alpha ) * cosf( tmp + tmp_1 );

		/* Loop over all samples */
		@@ -252,8 +265,83 @@ void ivas_param_ism_stereo_dmx(
		stereo_dmx[0][j] += cardioid_left[i] * tmp; /* DMX Left */
		stereo_dmx[1][j] += cardioid_right[i] * tmp; /* DMX Right */
		}
		#else
		if ( st_ivas->hSCE[0]->hCoreCoder[0]->ini_frame > 0 )
		{
		/* Smoothing */
		cardioid_left[i] = 0.75f * cardioid_left[i] + 0.25f * *( last_cardioid_left );
		grad = ( cardioid_left[i] - ( last_cardioid_left ) ) 2.0f / input_frame; /* for the right cardioid, multiply with -1 */
		/* Cardioids sum up to 1 */
		cardioid_right[i] = 1.0f - cardioid_left[i]; /* corresponds to: alpha + ( 1 - alpha ) * cosf( tmp + tmp_1 ); */
		/* Loop over all samples */
		for ( j = 0; j < input_frame / 2; j++ )
		{
		float last_cardioid_right;
		last_cardioid_right = 1.0f - *( last_cardioid_left );
		tmp = data[i][j];
		stereo_dmx[0][j] += ( ( ( last_cardioid_left ) + j grad ) * tmp ); /* DMX Left */
		stereo_dmx[1][j] += ( ( last_cardioid_right + j * grad * ( -1 ) ) * tmp ); /* DMX Right */
		}
		for ( ; j < input_frame; j++ )
		{
		tmp = data[i][j];
		stereo_dmx[0][j] += cardioid_left[i] * tmp; /* DMX Left */
		stereo_dmx[1][j] += cardioid_right[i] * tmp; /* DMX Right */
		ene_data += ( tmp * tmp ); /* energy of all objects combined */
		}
		}
		else
		{
		/* Cardioids sum up to 1 */
		cardioid_right[i] = 1.0f - cardioid_left[i]; /* corresponds to: alpha + ( 1 - alpha ) * cosf( tmp + tmp_1 ); */
		/* Loop over all samples */
		for ( j = 0; j < input_frame; j++ )
		{
		tmp = data[i][j];
		stereo_dmx[0][j] += cardioid_left[i] * tmp; /* DMX Left */
		stereo_dmx[1][j] += cardioid_right[i] * tmp; /* DMX Right */
		ene_data += ( tmp * tmp ); /* energy of all objects combined */
		}
		}
		*( last_cardioid_left ) = cardioid_left[i];
		last_cardioid_left++;
		#endif
		}

		#ifdef FIX_549_DMX_GAIN
		/* Energy compensation */
		for ( j = 0; j < input_frame; j++ )
		{
		ene_dmx += stereo_dmx[0][j] * stereo_dmx[0][j] + stereo_dmx[1][j] * stereo_dmx[1][j];
		}
		dmx_gain = sqrtf( ene_data / ( ene_dmx + EPSILON ) );
		/* Smoothing */
		if ( st_ivas->hSCE[0]->hCoreCoder[0]->ini_frame > 0 )
		{
		dmx_gain = 0.75f * dmx_gain + 0.25f * *( last_dmx_gain );
		/* 10ms ramp */
		grad = ( dmx_gain - ( last_dmx_gain ) ) 2.0f / input_frame; /* slope between two consecutive gains, 480 samples length */
		for ( i = 0; i < ( input_frame / 2 ); i++ )
		{
		stereo_dmx[0][i] = ( ( last_dmx_gain ) + i * grad );
		stereo_dmx[1][i] = ( ( last_dmx_gain ) + i * grad );
		}
		for ( ; i < input_frame; i++ )
		{
		stereo_dmx[0][i] *= dmx_gain;
		stereo_dmx[1][i] *= dmx_gain;
		}
		}
		else
		{
		for ( j = 0; j < input_frame; j++ )
		{
		stereo_dmx[0][j] *= dmx_gain;
		stereo_dmx[1][j] *= dmx_gain;
		}
		}
		*( last_dmx_gain ) = dmx_gain;
		#endif
		/* Copy the stereo dmx to data variable */
		mvr2r( stereo_dmx[0], data[0], input_frame );
		mvr2r( stereo_dmx[1], data[1], input_frame );