fix infinite CNG in ISM decoder when going to silence (cf1893bd) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_prot.h

+7 −0

Original line number	Diff line number	Diff line
		@@ -1163,6 +1163,13 @@ void ivas_get_ism_sid_quan_bitbudget(
		int16_t nBits_sce_id / o : number of Q bits for sce_id_dtx */
		);

		#ifdef FIX_ISM_DTX_INFINITE_CNG_ON_TRAILING_SILENCE
		void ivas_ism_dtx_limit_noise_energy_for_near_silence(
		SCE_DEC_HANDLE hSCE[],
		const int16_t sce_id_dtx,
		const int16_t nchan_transport
		);
		#endif

		/----------------------------------------------------------------------------------
		* DFT Stereo prototypes

lib_com/options.h

+1 −1

Original line number	Diff line number	Diff line
		@@ -209,6 +209,7 @@
		#define FIX_615_UBSAN_SPAR_TO_DIRAC /Dlb : Fix for UBSAN issue 615/
		#define FIX_626_VARIABLE_TYPE_MDCT_CONC /* FhG: trivial fix to fix USAN error */
		#define FIX_616_DIV_ZERO_MCT /FhG : Fix UBSAN division by zero error of issue 616/
		#define FIX_ISM_DTX_INFINITE_CNG_ON_TRAILING_SILENCE /* FhG: fix for cng in ISM DTX on sudden silence periods */

		/* ################## End BE DEVELOPMENT switches ######################### */

		@@ -220,7 +221,6 @@

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


		/* ################## End DEVELOPMENT switches ######################### */

		/* clang-format on */

lib_dec/ivas_dec.c

+21 −0

Original line number	Diff line number	Diff line
		@@ -180,6 +180,16 @@ ivas_error ivas_dec(
		{
		return error;
		}
		#ifdef FIX_ISM_DTX_INFINITE_CNG_ON_TRAILING_SILENCE

		/* decode dominant object first so we can limit the noise energy of the other objects */
		if ( ( error = ivas_sce_dec( st_ivas, st_ivas->hISMDTX.sce_id_dtx, &output[st_ivas->hISMDTX.sce_id_dtx], output_frame, nb_bits_metadata[st_ivas->hISMDTX.sce_id_dtx] ) ) )
		{
		return error;
		}

		ivas_ism_dtx_limit_noise_energy_for_near_silence( st_ivas->hSCE, st_ivas->hISMDTX.sce_id_dtx, st_ivas->nchan_transport );
		#endif
		}
		else if ( st_ivas->ism_mode == ISM_MODE_PARAM )
		{
		@@ -198,10 +208,21 @@ ivas_error ivas_dec(

		for ( n = 0; n < st_ivas->nchan_transport; n++ )
		{
		#ifdef FIX_ISM_DTX_INFINITE_CNG_ON_TRAILING_SILENCE
		/* for DTX frames, dominant object has already been decoded before */
		if ( !( ( ivas_total_brate == IVAS_SID_5k2 \|\| ivas_total_brate == FRAME_NO_DATA ) && n == st_ivas->hISMDTX.sce_id_dtx ) )
		{
		if ( ( error = ivas_sce_dec( st_ivas, n, &output[n], output_frame, nb_bits_metadata[n] ) ) != IVAS_ERR_OK )
		{
		return error;
		}
		}
		#else
		if ( ( error = ivas_sce_dec( st_ivas, n, &output[n], output_frame, nb_bits_metadata[n] ) ) != IVAS_ERR_OK )
		{
		return error;
		}
		#endif

		/* HP filtering */
		hp20( output[n], output_frame, st_ivas->mem_hp20_out[n], output_Fs );

lib_dec/ivas_ism_dtx_dec.c

+51 −0

Original line number	Diff line number	Diff line
		@@ -31,6 +31,7 @@
		*******************************************************************************************************/

		#include <stdint.h>
		#include <math.h>
		#include "options.h"
		#include "ivas_prot.h"
		#include "prot.h"
		@@ -156,3 +157,53 @@ ivas_error ivas_ism_dtx_dec(

		return IVAS_ERR_OK;
		}

		#ifdef FIX_ISM_DTX_INFINITE_CNG_ON_TRAILING_SILENCE
		/-------------------------------------------------------------------
		* ivs_ism_dtx_limit_noise_energy_for_near_silence
		*
		* for DTX frames where the energy of the sent noise estimate of the dominant object
		* is near silence, limit the other objects CNG energies to the same level
		*
		-------------------------------------------------------------------/
		void ivas_ism_dtx_limit_noise_energy_for_near_silence(
		SCE_DEC_HANDLE hSCE[],
		const int16_t sce_id_dtx,
		const int16_t nchan_transport )
		{
		float cng_noise_nrg_dominant;
		int16_t cng_noise_level_len;
		HANDLE_FD_CNG_COM hFdCngCom;

		hFdCngCom = hSCE[sce_id_dtx]->hCoreCoder[0]->hFdCngDec->hFdCngCom;
		cng_noise_level_len = hFdCngCom->stopFFTbin - hFdCngCom->startBand;
		cng_noise_nrg_dominant = dotp( hFdCngCom->cngNoiseLevel, hFdCngCom->cngNoiseLevel, cng_noise_level_len );

		if ( cng_noise_nrg_dominant < 1.f )
		{
		for ( int16_t n = 0; n < nchan_transport; n++ )
		{
		float cng_noise_nrg_obj;

		if ( n == sce_id_dtx )
		{
		continue;
		}

		hFdCngCom = hSCE[n]->hCoreCoder[0]->hFdCngDec->hFdCngCom;
		cng_noise_level_len = hFdCngCom->stopFFTbin - hFdCngCom->startBand;
		cng_noise_nrg_obj = dotp( hFdCngCom->cngNoiseLevel, hFdCngCom->cngNoiseLevel, cng_noise_level_len );

		if ( cng_noise_nrg_obj > cng_noise_nrg_dominant )
		{
		float fac;

		fac = sqrtf( cng_noise_nrg_dominant / cng_noise_nrg_obj );
		v_multc( hFdCngCom->cngNoiseLevel, fac, hFdCngCom->cngNoiseLevel, cng_noise_level_len );
		}
		}
		}

		return;
		}
		#endif

lib_enc/ivas_ism_dtx_enc.c

+0 −1

Original line number	Diff line number	Diff line
		@@ -312,7 +312,6 @@ void ivas_ism_get_sce_id_dtx(
		SCE_ENC_HANDLE hSCE[MAX_SCE], /* i/o: SCE encoder structure */
		const int16_t nchan_transport, /* i : number of transport channels */
		const int16_t input_frame /* i : input frame length per channel */

		)
		{
		float tmp_energy[MAX_NUM_OBJECTS];