Core coder functions: acelp_core, stere_tcx_core and other sub-functions (f64d3dd4) · Commits · SA4 / Audio / IVAS BASOP

lib_com/cnst.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -1425,6 +1425,7 @@ enum
		#define NB_LEADER 36
		#define NB_LDQ4 27
		#define FAC_LOG2 3.321928095f
		#define FAC_LOG2_BY10_Q16 21771

		#define NSV_MAX 34 /* maximal number of sub-vectors used by the AVQ */

lib_com/edct_fx.c

+1 −1

Original line number	Diff line number	Diff line
		@@ -746,7 +746,7 @@ void edxt_fx(
		move32();
		}

		y[sub( Nm1, imult1616( Nm1, shl( kernelType, 1 ) ) )] = L_shr( re[0], 1 );
		y[sub( Nm1, imult1616( Nm1, shr( kernelType, 1 ) ) )] = L_shr( re[0], 1 );
		move32();
		}
		ELSE /* inverse II = III */

lib_com/ivas_prot.h

+17 −4

Original line number	Diff line number	Diff line
		@@ -317,8 +317,8 @@ ivas_error pre_proc_ivas(
		const int16_t loc_harm, /* i : harmonicity flag */
		const float cor_map_sum, /* i : speech/music clasif. parameter */
		const int16_t vad_flag_dtx, /* i : HE-SAD flag with additional DTX HO */
		const float enerBuffer[CLDFB_NO_CHANNELS_MAX], /* i : energy buffer */
		const float fft_buff[2 * L_FFT], /* i : FFT buffer */
		/const/ float enerBuffer[CLDFB_NO_CHANNELS_MAX], /* i : energy buffer */
		/const/ float fft_buff[2 * L_FFT], /* i : FFT buffer */
		const int16_t MCT_flag, /* i : hMCT handle allocated (1) or not (0) */
		const int16_t vad_hover_flag, /* i : VAD hangover flag */
		const int16_t flag_16k_smc /* i : flag to indicate if the OL SMC is run at 16 kHz */
		@@ -635,6 +635,7 @@ ivas_error ivas_core_dec(
		#endif


		#ifndef IVAS_FLOAT_FIXED
		void encod_gen_2sbfr(
		Encoder_State st, / i/o: state structure */
		const float speech[], /* i : input speech */
		@@ -650,6 +651,7 @@ void encod_gen_2sbfr(
		const int16_t tdm_Pitch_reuse_flag, /* i : primary channel pitch reuse flag */
		const float tdm_Pri_pitch_buf[] /* i : pitch values for primary channel */
		);
		#endif

		void decod_gen_2sbfr(
		Decoder_State st, / i/o: decoder static memory */
		@@ -799,6 +801,7 @@ int16_t ivas_acelp_tcx20_switching(
		const int16_t flag_16k_smc /* i : flag to indicate if the OL SMC is run at 16 kHz */
		);

		#ifndef IVAS_FLOAT_FIXED
		void ivas_decision_matrix_enc(
		Encoder_State st, / i/o: encoder state structure */
		const int32_t element_brate, /* i : element bitrate */
		@@ -806,6 +809,16 @@ void ivas_decision_matrix_enc(
		const float enerBuffer[], /* i : energy buffer */
		const int16_t last_element_mode /* i : last element mode */
		);
		#else
		void ivas_decision_matrix_enc_fx(
		Encoder_State st, / i : encoder state structure */
		const Word32 element_brate, /* i : element bitrate */
		const Word16 fft_buff[], /* i : FFT buffer */
		const Word32 enerBuffer[], /* i : energy buffer */
		Word16 enerBuffer_exp,
		const Word16 last_element_mode /* i : last element mode */
		);
		#endif

		#ifdef IVAS_FLOAT_FIXED
		void ivas_signaling_enc_fx(
		@@ -8001,8 +8014,8 @@ void ivas_omasa_dirac_rend_jbm(
		#ifdef IVAS_FLOAT_FIXED
		void ivas_omasa_preProcessStereoTransportsForMovedObjects_fx(
		Decoder_Struct *st_ivas,
		Word32 inRe_fx[][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX],
		Word32 inIm_fx[][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX],
		Word32 inRe_fx[][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX], /cldfb_buf_q/
		Word32 inIm_fx[][CLDFB_SLOTS_PER_SUBFRAME][CLDFB_NO_CHANNELS_MAX], /cldfb_buf_q/
		Word16 *cldfb_buf_q,
		const Word16 nBins,
		const Word16 subframe

lib_com/ivas_prot_fx.h

+78 −11

Original line number	Diff line number	Diff line
		@@ -2653,6 +2653,14 @@ void ivas_param_mc_metadata_open_fx(
		HANDLE_IVAS_PARAM_MC_METADATA hMetadataPMC /* o : handle for the Parametric MC parameter coding state */
		);

		Word16 mdct_classifier_ivas_fx(
		Encoder_State st, / i/o: Encoder state variable */
		const Word16 fft_buff, / i : FFT spectrum from fft_rel */
		const Word32 enerBuffer[], /* i : energy buffer */
		Word16 enerBuffer_exp, /* i: enenrgy buffer exponent */
		const Word32 brate /* i : current brate, IVAS: nominal bitrate, EVS: st->total_brate */
		);

		/----------------------------------------------------------------------------------
		* Range Coder prototypes
		----------------------------------------------------------------------------------/
		@@ -2723,6 +2731,38 @@ void core_switching_pre_enc_ivas_fx(
		const Word16 active_cnt, /* i : active frame counter */
		const Word16 last_element_mode /* i : last_element_mode */
		);

		void encod_gen_2sbfr(
		Encoder_State st, / i/o: state structure */
		const Word16 speech[], /* i : input speech */
		const Word16 Aw[], /* i : weighted A(z) unquantized for subframes */
		const Word16 Aq[], /* i : LP coefficients */
		const Word16 res, / i : residual signal */
		Word16 syn, / i/o: core synthesis */
		Word16 exc, / i/o: current non-enhanced excitation */
		Word16 exc2, / i/o: current enhanced excitation */
		Word16 pitch_buf, / i/o: floating pitch values for each subframe */
		Word16 voice_factors, / o : voicing factors */
		Word16 bwe_exc, / o : excitation for SWB TBE */
		const Word16 tdm_Pitch_reuse_flag, /* i : primary channel pitch reuse flag */
		const Word16 tdm_Pri_pitch_buf[], /* i : pitch values for primary channel */
		Word16 *Q_new );

		void acelp_fast_fx(
		BSTR_ENC_HANDLE hBstr, /* i/o: encoder bitstream handle */
		const Word16 cdk_index, /* i : codebook index */
		const Word16 dn_orig[L_SUBFR],
		/* i : corr. between target and h[]. */ // Q_new + 1
		Word16 Q_dn,
		const Word16 cn[L_SUBFR],
		/* i : residual after long term prediction */ // Q_new + 1
		const Word16 H[L_SUBFR],
		/* i : impulse response of weighted synthesis filter */ // e(norm_s(H[0])+1)
		Word16 code[L_SUBFR], /* o : algebraic (fixed) codebook excitation */
		Word16 y[], /* o : filtered fixed codebook excitation */
		const Word16 L_subfr /* i : subframe length */
		);

		#endif

		void ivas_mdct_quant_coder_fx(
		@@ -2741,15 +2781,42 @@ void ivas_mdct_tcx10_bit_distribution_fx(

		void QuantizeSpectrum_ivas_fx(
		Encoder_State st, / i/o: encoder state structure */
		const float A[], /* i : quantized coefficients NxAz_q[M+1] */
		const Word16 A_fx[], /* i : quantized coefficients NxAz_q[M+1] */
		const Word16 Aqind[], /* i : frame-independent quantized coefficients (M+1) */
		float gainlpc[], /* i : MDCT gains of the previous frame */
		float synth[], /* o : synthesis buffer */
		const int16_t nb_bits, /* i : bit budget */
		const int16_t tnsSize, /* i : number of tns parameters put into prm */
		int16_t prm[], /* o : tcx parameters */
		const int16_t frame_cnt, /* i : frame counter in the super_frame */
		Word16 gainlpc_fx[], /* i : MDCT gains of the previous frame */
		Word16 gainlpc_e[], /* i : exponents of MDCT gains of the previous frame */
		Word16 synth[], /* o : synthesis buffer, Q0 */
		const Word16 nb_bits, /* i : bit budget */
		const Word16 tnsSize, /* i : number of tns parameters put into prm */
		Word16 prm[], /* o : tcx parameters */
		const Word16 frame_cnt, /* i : frame counter in the super_frame */
		CONTEXT_HM_CONFIG hm_cfg, / i : HM configuration */
		const int16_t vad_hover_flag /* i : VAD hangover flag */
		const Word16 vad_hover_flag /* i : VAD hangover flag */
		);

		void InternalTCXDecoder_fx(
		Encoder_State st, / i/o: state handle */
		const Word16 frame_cnt, /* i : frame counter in the super_frame */
		const Word16 L_frameTCX, /* i : full frame length */
		const Word16 L_frame, /* i : frame length */
		const Word16 L_spec, /* i : length of the coded spectrum */
		const Word16 tcx_offset, /* i : folding point offset relative to the end of the previous frame */
		const Word16 noiseFillingBorder, /* i : noise filling border */
		const Word32 x_quant_fx, / i : quantized spectrum */
		const Word32 ener_fx, /* i : energy of the quantized spectrum */
		const Word16 ener_e, /* i : energy of the quantized spectrum exponent */
		Word16 lf_deemph_fact_fx[], /* i/o: low frequency deemphasis factors */
		const Word16 fac_ns_fx, /* i : noise filling level, Q15 */
		const Word16 nf_seed, /* i : noise filling random seed, Q0 */
		const Word16 A_fx, / i : LPC representation of the FDNS gains */
		Word16 gainlpc_fx[], /* i/o: FDNS gains */
		Word16 gainlpc_e[], /* i/o: FDNS gains exponent */
		const Word16 hm_active, /* i : flag indicating if the harmonic model is active */
		Word16 gain_tcx_fx, /* i/o: global gain / quantized global gain */
		Word16 gain_tcx_e, / i/o: global gain / quantized global gain exponent */
		Word32 spectrum_fx[], /* o : dequantized spectrum */
		Word16 spectrum_e, / o : dequantized spectrum */
		Word16 synth[], /* o : time domain signal */
		Word16 gain_tcx_q / o : quantized global gain (at low bitrates) */
		);
		#endif

lib_com/prot_fx.h

+39 −0

Original line number	Diff line number	Diff line
		@@ -1746,6 +1746,17 @@ void wtda_fx(
		const Word16 L /* i : length */
		);

		#ifdef IVAS_FLOAT_FIXED
		void wtda_ext_fx(
		const Word16 new_audio, / i : input audio (Q_in) */
		Word16 wtda_audio, / o : windowed audio (Q_in) */
		const Word16 left_mode, /* i : window overlap of previous frame (0: full, 2: none, or 3: half) */
		const Word16 right_mode, /* i : window overlap of current frame (0: full, 2: none, or 3: half) */
		const Word16 L, /* i : length */
		const UWord16 kernel_type /* i : transform kernel type (0 - 3) */
		);
		#endif

		/*========================================================================================================/
		swb_bwe_com_lr_fx.c
		/========================================================================================================*/
		@@ -3947,6 +3958,19 @@ void mdct_window_aldo(
		// tcx_utils.c
		Word16 getInvFrameLen( const Word16 L_frame ); /* returns 1/L_frame in Q21 format */

		#ifdef IVAS_FLOAT_FIXED
		void tcx_get_windows(
		TCX_CONFIG_HANDLE hTcxCfg, /* i : TCX configuration */
		const Word16 left_mode, /* i: overlap mode of left window half */
		const Word16 right_mode, /* i: overlap mode of right window half */
		Word16 left_overlap, / o: left overlap length */
		const PWord16 *left_win, / o: left overlap window */
		Word16 right_overlap, / o: right overlap length */
		const PWord16 *right_win, / o: right overlap window */
		const Word8 fullband /* i: fullband flag */
		);
		#endif

		void WindowSignal(
		TCX_CONFIG_HANDLE hTcxCfg, /* i : configuration of TCX */
		Word16 offset, /* i : left folding point offset relative to the i signal pointer */
		@@ -7479,6 +7503,7 @@ void HQ_nbfec_init_fx(
		HQ_NBFEC_HANDLE hHQ_nbfec /* i/o: HQ NB FEC data handle */
		);

		void GetAttackForTCXDecision_fx( Word32 const pSubblockNrg, Word32 const pAccSubblockNrg, Word16 nSubblocks, Word16 nPastSubblocks, Word16 attackRatioThreshold, Word16 pbIsAttackPresent, Word16 pAttackIndex );

		// FEC_HQ_phase_ecu_fx.c
		void hq_ecu_fx(
		@@ -10770,4 +10795,18 @@ Word16 msvq_stage1_dct_recalc_candidates_fdcng_wb_fx(
		Word32 dist_ptr_fx, / i/o: updated MSE vector for stage1 */
		Word16 dist_ptr_e / i/o: exp for updated MSE vector for stage1 */
		);

		void FEC_encode_ivas_fx(
		BSTR_ENC_HANDLE hBstr, /* i/o: encoder bitstream handle */
		const ACELP_config acelp_cfg, /* i/o: configuration of the ACELP */
		const Word16 synth, / i : pointer to synthesized speech for E computation */
		const Word16 coder_type, /* i : type of coder */
		Word16 clas, /* i : signal clas for current frame */
		const Word16 fpit, / i : close loop fractional pitch buffer Q6 */
		const Word16 res, / i : LP residual signal frame */
		Word16 last_pulse_pos, / i/o: Position of the last pulse */
		const Word16 L_frame, /* i : Frame length */
		const Word32 total_brate, /* i : total codec bitrate */
		const Word16 Q_synth /* i : input scaling */
		);
		#endif