Merge branch 'ericsson/low-rate-encoding-of-transients' into 'main' (27451d34) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_prot.h

+15 −1

Original line number	Diff line number	Diff line
		@@ -215,8 +215,14 @@ ivas_error pre_proc_front_ivas(
		const int16_t flag_16k_smc, /* i : flag to indicate if the OL SMC is run at 16 kHz */
		const int16_t front_vad_flag, /* i : front-VAD flag to overwrite VAD decision */
		const int16_t force_front_vad, /* i : flag to force VAD decision */
		#ifdef LOW_RATE_TRANS
		const int16_t front_vad_dtx_flag, /* i : front-VAD DTX flag to overwrite VAD decision*/
		const int32_t ivas_total_brate, /* i : IVAS total bitrate */
		const int16_t ivas_format /* i : IVAS format */
		#else
		const int16_t front_vad_dtx_flag /* i : front-VAD DTX flag to overwrite VAD decision*/
		,const int32_t ivas_total_brate /* i : IVAS total bitrate */
		#endif
		);

		ivas_error pre_proc_ivas(
		@@ -621,6 +627,14 @@ void set_transient_stereo(
		float currFlatness[] /* i/o: current flatness */
		);

		#ifdef LOW_RATE_TRANS
		int16_t transient_analysis(
		TRAN_DET_HANDLE hTranDet, /* i : handle transient detection */
		const float cor_map_LT[], /* i : LT correlation map */
		const float multi_harm_limit /* i : multi harminic threshold */
		);
		#endif

		void ivas_post_proc(
		SCE_DEC_HANDLE hSCE, /* i/o: SCE decoder structure */
		CPE_DEC_HANDLE hCPE, /* i/o: CPE decoder structure */

lib_com/options.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -155,6 +155,7 @@
		#define FIX_I59_DELAY_ROUNDING /* Issue 59: rounding in sample domain instead of nanosec for IVAS_ENC_GetDelay() and IVAS_DEC_GetDelay() */
		#define FIX_FIX_I59 /* Issue 59: small fix concerning LFE delay rounding */
		#define FIX_I59_CREND /* Issue 59: Fixes for gcc compiler warnings in ivas_crend_utest_utils.c */
		#define LOW_RATE_TRANS /* Eri: Contribution 20: low rate encoding of transients */

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

lib_enc/ivas_core_pre_proc_front.c

+42 −2

Original line number	Diff line number	Diff line
		@@ -45,6 +45,14 @@
		#include <math.h>


		#ifdef LOW_RATE_TRANS
		/---------------------------------------------------------------
		* Local constants
		---------------------------------------------------------------/

		#define SCE_SMC_THR 16000
		#endif

		/-------------------------------------------------------------------
		* Local function prototypes
		--------------------------------------------------------------------/
		@@ -100,9 +108,15 @@ ivas_error pre_proc_front_ivas(
		const int16_t flag_16k_smc, /* i : flag to indicate if the OL SMC is run at 16 kHz */
		const int16_t front_vad_flag, /* i : front-VAD flag to overwrite VAD decision */
		const int16_t force_front_vad, /* i : flag to force VAD decision */
		#ifdef LOW_RATE_TRANS
		const int16_t front_vad_dtx_flag, /* i : front-VAD DTX flag to overwrite VAD decision*/
		const int32_t ivas_total_brate, /* i : IVAS total bitrate - for setting the DTX */
		const int16_t ivas_format /* i : IVAS format */
		#else
		const int16_t front_vad_dtx_flag /* i : front-VAD DTX flag to overwrite VAD decision*/
		,
		const int32_t ivas_total_brate /* i : IVAS total bitrate - for setting the DTX */
		#endif
		)
		{
		float inp_12k8, new_inp_12k8; /* pointers to current frame and new data */
		@@ -784,7 +798,11 @@ ivas_error pre_proc_front_ivas(
		else if ( element_mode != IVAS_CPE_MDCT )
		{
		/* SNR-based speech/music classification */
		#ifdef LOW_RATE_TRANS
		if ( ( element_mode >= IVAS_CPE_DFT && element_brate >= IVAS_24k4 ) \|\| ( element_mode == IVAS_SCE && element_brate >= SCE_SMC_THR ) )
		#else
		if ( ( element_mode >= IVAS_CPE_DFT && element_brate >= IVAS_24k4 ) \|\| ( element_mode == IVAS_SCE && element_brate >= 16000 ) )
		#endif
		{
		if ( flag_16k_smc )
		{
		@@ -802,7 +820,29 @@ ivas_error pre_proc_front_ivas(
		smc_dec = ivas_acelp_tcx20_switching( st, inp_12k8, wsp, non_staX, pitch_fr, voicing_fr, currFlatness, lsp_mid, stab_fac, res_cod_SNR_M, flag_16k_smc );
		}
		}

		#ifdef LOW_RATE_TRANS
		/* Switch to ACELP for non-harmonic transient signals */
		else if ( ( ( ivas_format == STEREO_FORMAT && element_brate <= IVAS_16k4 ) \|\| ( ivas_format == ISM_FORMAT && element_brate < SCE_SMC_THR ) ) && ( loc_harm[0] != 1 ) && smc_dec == MUSIC )
		{
		if ( element_mode == IVAS_SCE )
		{
		if ( transient_analysis( st->hTranDet, st->hNoiseEst->cor_map, st->hNoiseEst->multi_harm_limit ) )
		{
		smc_dec = SPEECH;
		}
		}
		else if ( element_mode == IVAS_CPE_DFT )
		{
		for ( i = 0; i < CPE_CHANNELS; i++ )
		{
		if ( smc_dec != SPEECH && transient_analysis( hCPE->hCoreCoder[i]->hTranDet, st->hNoiseEst->cor_map, st->hNoiseEst->multi_harm_limit ) )
		{
		smc_dec = SPEECH; /* overwrite initial music decision, intial SPEECH_MUSIC never changed */
		}
		}
		}
		}
		#endif
		/* 2nd stage speech/music classification (ACELP/GSC/TCX core selection) */
		ivas_smc_mode_selection( st, element_brate, smc_dec, relE, Etot, attack_flag, inp_12k8, S_map, flag_spitch );

lib_enc/ivas_cpe_enc.c

+7 −0

Original line number	Diff line number	Diff line
		@@ -426,10 +426,17 @@ ivas_error ivas_cpe_enc(

		for ( n = 0; n < n_CoreChannels; n++ )
		{
		#ifdef LOW_RATE_TRANS
		error = pre_proc_front_ivas( NULL, hCPE, hCPE->element_brate, nb_bits_metadata, input_frame, n, old_inp_12k8[n], old_inp_16k[n], &Etot[n], &ener[n], &relE[n], A[n], Aw[n], epsP[n], lsp_new[n], lsp_mid[n],
		&vad_hover_flag[n], &attack_flag[n], realBuffer[n], imagBuffer[n], old_wsp[n], pitch_fr[n], voicing_fr[n], &loc_harm[n], &cor_map_sum[n], &vad_flag_dtx[n], enerBuffer[n],
		fft_buff[n], A[0], lsp_new[0], currFlatness[n], tdm_ratio_idx, fr_bands, Etot_LR, lf_E, localVAD_HE_SAD, band_energies_LR, 0, st_ivas->hSpar != NULL ? st_ivas->hSpar->front_vad_flag : 0, 0, 0,
		ivas_total_brate, st_ivas->hEncoderConfig->ivas_format );
		#else
		error = pre_proc_front_ivas( NULL, hCPE, hCPE->element_brate, nb_bits_metadata, input_frame, n, old_inp_12k8[n], old_inp_16k[n], &Etot[n], &ener[n], &relE[n], A[n], Aw[n], epsP[n], lsp_new[n], lsp_mid[n],
		&vad_hover_flag[n], &attack_flag[n], realBuffer[n], imagBuffer[n], old_wsp[n], pitch_fr[n], voicing_fr[n], &loc_harm[n], &cor_map_sum[n], &vad_flag_dtx[n], enerBuffer[n],
		fft_buff[n], A[0], lsp_new[0], currFlatness[n], tdm_ratio_idx, fr_bands, Etot_LR, lf_E, localVAD_HE_SAD, band_energies_LR, 0, st_ivas->hSpar != NULL ? st_ivas->hSpar->front_vad_flag : 0, 0, 0,
		ivas_total_brate );
		#endif
		if ( error != IVAS_ERR_OK )
		{
		return error;

lib_enc/ivas_ism_enc.c

+8 −0

Original line number	Diff line number	Diff line
		@@ -147,11 +147,19 @@ ivas_error ivas_ism_enc(
		* Front Pre-processing
		----------------------------------------------------------------/

		#ifdef LOW_RATE_TRANS
		error = pre_proc_front_ivas( hSCE, NULL, hSCE->element_brate, nb_bits_metadata[sce_id], input_frame, 0, old_inp_12k8[sce_id][0], old_inp_16k[sce_id][0],
		&Etot[sce_id][0], &ener[sce_id][0], &relE[sce_id][0], A[sce_id][0], Aw[sce_id][0], epsP[sce_id][0], lsp_new[sce_id][0], lsp_mid[sce_id][0],
		&vad_hover_flag[sce_id][0], &attack_flag[sce_id][0], realBuffer[sce_id][0], imagBuffer[sce_id][0], old_wsp[sce_id][0], pitch_fr[sce_id][0], voicing_fr[sce_id][0], &loc_harm[sce_id][0], &cor_map_sum[sce_id][0], &vad_flag_dtx[sce_id][0], enerBuffer[sce_id][0],
		fft_buff[sce_id][0], A[sce_id][0], lsp_new[sce_id][0], currFlatness[0], 0, fr_bands, Etot_LR, lf_E, localVAD_HE_SAD, NULL, 0, 0, 0, 0,
		st_ivas->hEncoderConfig->ivas_format, st_ivas->hEncoderConfig->ivas_total_brate );
		#else
		error = pre_proc_front_ivas( hSCE, NULL, hSCE->element_brate, nb_bits_metadata[sce_id], input_frame, 0, old_inp_12k8[sce_id][0], old_inp_16k[sce_id][0],
		&Etot[sce_id][0], &ener[sce_id][0], &relE[sce_id][0], A[sce_id][0], Aw[sce_id][0], epsP[sce_id][0], lsp_new[sce_id][0], lsp_mid[sce_id][0],
		&vad_hover_flag[sce_id][0], &attack_flag[sce_id][0], realBuffer[sce_id][0], imagBuffer[sce_id][0], old_wsp[sce_id][0], pitch_fr[sce_id][0], voicing_fr[sce_id][0], &loc_harm[sce_id][0], &cor_map_sum[sce_id][0], &vad_flag_dtx[sce_id][0], enerBuffer[sce_id][0],
		fft_buff[sce_id][0], A[sce_id][0], lsp_new[sce_id][0], currFlatness[0], 0, fr_bands, Etot_LR, lf_E, localVAD_HE_SAD, NULL, 0, 0, 0, 0,
		st_ivas->hEncoderConfig->ivas_total_brate );
		#endif
		if ( error != IVAS_ERR_OK )
		{
		return error;