commit 1 (2c9d5e6f) · Commits · SA4 / Audio / IVAS BASOP

lib_com/cnst.h

+11 −0

Original line number	Diff line number	Diff line
		@@ -209,6 +209,15 @@ enum{
		#define L_FRAME8k 160 /* Frame size in samples at 8kHz */
		#define L_FRAME4k 80 /* Frame size in samples at 4kHz */

		#define ONE_BY_L_FRAME48k_Q31 2236962
		#define ONE_BY_L_FRAME32k_Q31 3355443
		#define ONE_BY_L_FRAME25_6k_Q31 4194304
		#define ONE_BY_L_FRAME16k_Q31 6710886
		#define ONE_BY_L_FRAME_Q31 8388608
		#define ONE_BY_240_Q31 8947849
		#define ONE_BY_L_FRAME8k_Q31 13421772
		#define ONE_BY_L_FRAME4k_Q31 26843546

		#define L_SPEC48k 800 /* HQ spectrum length at 48kHz */
		#define L_SPEC32k 640 /* HQ spectrum length at 32kHz */
		#define L_SPEC16k 320 /* HQ spectrum length at 16kHz */
		@@ -737,6 +746,7 @@ enum
		#define PIT_L_INTERPOL6_2 17
		#define PIT_FIR_SIZE6_2 ( PIT_UP_SAMP6 * PIT_L_INTERPOL6_2 + 1 )
		#define E_MIN 0.0035f /* minimum allowable energy */
		#define E_MIN_Q11_FX 7 /* minimum allowable energy in Q11*/
		#define STEP_DELTA 0.0625f /* quantization step for tilt compensation of gaussian cb. excitation */
		#define GAMMA_EV 0.92f /* weighting factor for core synthesis error weighting */
		#define FORMANT_SHARPENING_NOISE_THRESHOLD 21.0f /* lp_noise level above which formant sharpening is deactivated */
		@@ -2752,6 +2762,7 @@ enum
		#define Fs_2_16k 20480 /* lsf max value (Use in reorder_fx.c) */

		#define LG10 24660 /* 10log10(2) in Q13 /
		#define TEN_MULT_LOG10_2_IN_Q29 1616142464 /* (10 * log10(2)) = 3.0103 (in Q29) */
		#define LG10_s3_0 16440 /* 10log10(2)/1.55 = 1.00343331 in Q14 /
		#define LOG2_10 27213 /* log base 2 of 10 in Q12 */
		#define LOG2_10_Q29 1783446566 /* log base 2 of 10 in Q12 */

lib_com/ivas_prot.h

+38 −3

Original line number	Diff line number	Diff line
		@@ -703,6 +703,7 @@ void stereo_tcx_init_enc(
		Encoder_State st / i/o: encoder state structure */
		);

		#ifndef IVAS_FLOAT_FIXED
		void stereo_tcx_core_enc(
		Encoder_State st, / i/o: encoder state structure */
		const float new_samples_12k8[], /* i : buffer of input signal @12.8 kHz */
		@@ -715,7 +716,6 @@ void stereo_tcx_core_enc(
		const int16_t vad_hover_flag /* i : VAD hangover flag */
		);

		#ifndef IVAS_FLOAT_FIXED
		void stereo_tcx_core_dec(
		Decoder_State st, / i/o: decoder state structure */
		const FRAME_MODE frameMode, /* i : Decoder frame mode */
		@@ -1805,10 +1805,12 @@ void stereo_dft_enc_res(
		const int16_t max_bits
		);

		#ifndef IVAS_FLOAT_FIXED
		void stereo_dft_enc_write_BS(
		CPE_ENC_HANDLE hCPE, /* i/o: CPE encoder structure */
		int16_t nb_bits / o : number of bits written */
		);
		#endif

		void stereo_dtf_cng(
		CPE_DEC_HANDLE hCPE, /* i/o: CPE decoder structure */
		@@ -1857,6 +1859,7 @@ void stereo_dft_dequantize_itd(
		const int32_t output_Fs
		);

		#ifndef IVAS_FLOAT_FIXED
		void stereo_dft_enc_sid_calc_coh(
		STEREO_DFT_ENC_DATA_HANDLE hStereoDft, /* i/o: DFT stereo handle */
		float prev_cohBand[2(STEREO_DFT_BAND_MAX/2)], / i/o: Previous coherence */
		@@ -1864,7 +1867,9 @@ void stereo_dft_enc_sid_calc_coh(
		int16_t first_SID, / i/o: First SID indicator */
		float cohBand / i/o: Coherence per band */
		);
		#endif

		#ifndef IVAS_FLOAT_FIXED
		void stereo_dft_enc_sid_coh(
		BSTR_ENC_HANDLE hBstr, /* i/o: bitstream handle */
		float mem_cohBand, / i/o: Coherence memory */
		@@ -1872,6 +1877,7 @@ void stereo_dft_enc_sid_coh(
		int16_t nb_bits, / i/o: number of bits written */
		float cohBand / i/o: Coherence per band */
		);
		#endif
		#ifdef IVAS_FLOAT_FIXED
		void stereo_dft_dec_sid_coh_fx(
		Decoder_State st, / i/o: decoder state structure */
		@@ -3102,7 +3108,17 @@ void TNSAnalysisStereo(
		int16_t param_core[][NB_DIV * NPRM_DIV], /* o : quantized noise filling level */
		const int16_t mct_on /* i : flag mct block (1) or stereo (0) */
		);

		#ifdef IVAS_FLOAT_FIXED
		void TNSAnalysisStereo_fx(
		Encoder_State *sts, / i : encoder state handle */
		Word32 mdst_spectrum_fx[CPE_CHANNELS][NB_DIV], / o : MDST spectrum Qx*/
		const Word16 bWhitenedDomain, /* i : whitened domain flag Q0*/
		Word16 tnsSize[CPE_CHANNELS][NB_DIV], /* i : number of tns parameters put into prm Q0*/
		Word16 tnsBits[CPE_CHANNELS][NB_DIV], /* i : number of tns bits in the frame Q0*/
		Word16 param_core[][NB_DIV * NPRM_DIV], /* o : TNS parameters Q0*/
		const Word16 mct_on /* i : flag mct block (1) or stereo (0) Q0*/
		);
		#endif
		void InternalTCXDecoder(
		Encoder_State st, / i/o: state handle */
		const int16_t frame_cnt, /* i : frame counter in the super_frame */
		@@ -3681,7 +3697,26 @@ ivas_error front_vad(
		float PS_out, / o : energy spectrum */
		float Bin_E_out / o : log-energy spectrum of the current frame*/
		);

		#ifdef IVAS_FLOAT_FIXED
		ivas_error front_vad_fx(
		CPE_ENC_HANDLE hCPE, /* i/o: CPE encoder structure, nullable */
		Encoder_State st, / i/o: encoder state structure */
		const ENCODER_CONFIG_HANDLE hEncoderConfig, /* i : configuration structure */
		FRONT_VAD_ENC_HANDLE hFrontVads, / i/o: front-VAD handles */
		const int16_t MCT_flag, /* i : hMCT handle allocated (1) or not (0) */
		const int16_t input_frame, /* i : frame length */
		int16_t vad_flag_dtx[], /* o : HE-SAD flag with additional DTX HO */
		float fr_bands[][2 * NB_BANDS], /* i : energy in frequency bands */
		float Etot_LR[], /* o : total energy Left & Right channel */
		float lf_E[][2 * VOIC_BINS], /* i : per bin spectrum energy in lf, LR channels */
		int16_t localVAD_HE_SAD[], /* o : HE-SAD flag without hangover, LR channels */
		int16_t vad_hover_flag[], /* o : VAD hangover flag */
		float band_energies_LR[2 * NB_BANDS], /* o : energy in critical bands without minimum noise floor E_MIN */
		float PS_out, / o : energy spectrum */
		float Bin_E_out, / o : log-energy spectrum of the current frame*/
		Word16 Qinp
		);
		#endif
		ivas_error front_vad_spar(
		SPAR_ENC_HANDLE hSpar, /* i/o: SPAR encoder structure */
		const float omni_in, / i : omnidirectional input signal */

lib_com/ivas_prot_fx.h

+71 −5

Original line number	Diff line number	Diff line
		@@ -296,10 +296,10 @@ Word16 res_bpf_adapt_fx(

		// ivas_sns_com_fx.c
		void sns_compute_scf_fx(
		Word32 spectrum[],
		Word32 spectrum[], /* i : Spectrum (Q_in) */
		const PsychoacousticParameters *pPsychParams,
		const Word16 L_frame,
		Word32 *scf,
		Word32 scf, / o : Scalefactors (Q16)*/
		Word16 q );

		void sns_interpolate_scalefactors_fx(
		@@ -1402,10 +1402,18 @@ void dtx_read_padding_bits_fx(
		DEC_CORE_HANDLE st,
		const Word16 num_bits );

		void FdCngEncodeMDCTStereoSID_fx(
		CPE_ENC_HANDLE hCPE /* i/o: CPE encoder state structure */
		);

		void FdCngDecodeMDCTStereoSID_fx(
		CPE_DEC_HANDLE hCPE /* i/o: CPE decoder state structure */
		);

		void FdCngEncodeDiracMDCTStereoSID_fx(
		CPE_ENC_HANDLE hCPE /* i/o: CPE encoder state structure */
		);

		void FdCngDecodeDiracMDCTStereoSID_fx(
		CPE_DEC_HANDLE hCPE /* i/o: CPE decoder state structure */
		);
		@@ -2700,7 +2708,6 @@ void rc_uni_enc_encode_bits(
		const Word16 bits /* i : Number of bits */
		);

		#ifdef IVAS_FLOAT_FIXED
		void stereo_dft_enc_process_fx(
		CPE_ENC_HANDLE hCPE, /* i/o: CPE encoder structure */
		const Word16 vad_flag_dtx[], /* i : VAD dtx flags */
		@@ -2708,6 +2715,54 @@ void stereo_dft_enc_process_fx(
		const Word16 input_frame /* i : input frame length */
		);

		void stereo_dft_cng_side_gain_fx(
		STEREO_DFT_ENC_DATA_HANDLE hStereoDft, /* i/o: DFT stereo handle */
		STEREO_CNG_ENC_HANDLE hStereoCng, /* i/o: Stereo CNG data structure */
		const Word32 core_brate, /* i : core bitrate */
		const Word32 last_core_brate, /* i : last core bitrate */
		const Word16 bwidth /* i : audio band-width */
		);

		Word16 quantize_sns_fx(
		Word32 sns_in_fx[CPE_CHANNELS][NB_DIV][M], /* sns_e */
		Word32 snsQ_out_fx[CPE_CHANNELS][NB_DIV][M], /* sns_e */
		Word16 *sns_e,
		Encoder_State **sts,
		Word16 indices, / Q0 */
		Word16 zero_side_flag, / Q0 */
		Word16 sns_stereo_mode ); / Q0 */

		void stereo_cng_upd_counters_fx(
		STEREO_CNG_ENC_HANDLE hStereoCng, /* i/o: Stereo CNG data structure */
		const Word32 element_mode, /* i : element mode */
		const Word16 nbands, /* i : Number of bands in active */
		const Word32 sidSideGain[], /* i : SID side gains */
		const Word16 burst_ho_count, /* i : Hang-over count */
		Word16 coh_fade_counter / i : Coherence fade counter */
		);


		void stereo_dft_enc_sid_calc_coh_fx(
		STEREO_DFT_ENC_DATA_HANDLE hStereoDft, /* i/o: DFT stereo handle */
		Word32 prev_cohBand[2 * (STEREO_DFT_BAND_MAX / 2)], /* i/o: Previous coherence Q31 */
		Word16 td_active, / i/o: TD stereo mode indicator */
		Word16 first_SID, / i/o: First SID indicator */
		Word32 cohBand / i/o: Coherence per band Q31 */
		);

		void stereo_dft_enc_sid_coh_fx(
		BSTR_ENC_HANDLE hBstr, /* i/o: bitstream handle */
		Word32 mem_cohBand, / i/o: Coherence memory */
		const Word16 nbands, /* i : number of DFT stereo bands */
		Word16 nb_bits, / i/o: number of bits written */
		Word32 cohBand / i/o: Coherence per band */
		);

		void stereo_dft_enc_write_BS_fx(
		CPE_ENC_HANDLE hCPE, /* i/o: CPE encoder structure */
		Word16 nb_bits / o : number of bits written */
		);

		void spectral_balancer_fx16(
		Word16 signal, / i/o : signal Qx */
		Word16 mem, / i/o : mem Qx */
		@@ -2763,8 +2818,6 @@ void acelp_fast_fx(
		const Word16 L_subfr /* i : subframe length */
		);

		#endif

		void ivas_mdct_quant_coder_fx(
		CPE_ENC_HANDLE hCPE, /* i/o: Encoder CPE handle */
		Word16 tnsBits[CPE_CHANNELS][NB_DIV], /* i : bits needed for TNS parameters */
		@@ -2819,4 +2872,17 @@ void InternalTCXDecoder_fx(
		Word16 synth[], /* o : time domain signal */
		Word16 gain_tcx_q / o : quantized global gain (at low bitrates) */
		);

		void stereo_tcx_core_enc(
		Encoder_State st, / i/o: encoder state structure */
		const float new_samples_12k8[], /* i : buffer of input signal @12.8 kHz */
		const float new_samples_16k[], /* i : buffer of input signal @16 kHz */
		const Word16 Aw_fx[], /* i : weighted A(z) unquant. for subframes, Q12 */
		float lsp_new[], /* i : LSPs at the end of the frame, Q15 */
		float lsp_mid[], /* i : LSPs in the middle of the frame, Q15 */
		Word16 pitch_buf_fx[NB_SUBFR16k], /* o : pitch for each subframe, Q6 */
		const Word16 last_element_mode, /* i : last element mode, Q0 */
		const Word16 vad_hover_flag /* i : VAD hangover flag, Q0 */
		);

		#endif

lib_com/ivas_rom_com.c

+8 −0

Original line number	Diff line number	Diff line
		@@ -962,6 +962,14 @@ const float nf_tw_smoothing_coeffs[N_LTP_GAIN_MEMS] =
		* Stereo DTX tables
		----------------------------------------------------------------------------------/

		const Word32 dft_cng_coh_pred_Q29[STEREO_DFT_N_COH_PRED][STEREO_DFT_COH_PRED_COEFFS] =
		{
		{ 322840352, -54249732, 681323520, 45657648, 149804160, 359663232, 43794172, 101391832, -92841088, 172591648, 5010079, -45280228, 194478800, -229067248, 407153760 },
		{ 459212000, 225880928, 221880688, -5028333, -214586768, 743996800, 6672231, -8234526, 84004728, 277797408, 86032488, -163012256, 256499744, -301048768, 742268608 },
		{ 108739448, 26209502, 141415552, 12636868, 49403400, 212415664, 7013681, -5973762, 19040664, 524015552, 5351529, 87486872, 205720880, -196222560, 1360794368 },
		{ 748217664, -70545376, 531195648, 396387904, 367147232, 900332, -121036472, -6527276, 582294464, -61989264, 12010339, -24459838, -180512112, 226957888, 267986096 }
		};

		const float dft_cng_coh_pred[STEREO_DFT_N_COH_PRED][STEREO_DFT_COH_PRED_COEFFS] =
		{
		{ 0.601337f, -0.101048f, 1.269064f, 0.085044f, 0.279032f, 0.669925f, 0.081573f, 0.188857f, -0.172930f, 0.321477f, 0.009332f, -0.084341f, 0.362245f, -0.426671f, 0.758383f},

lib_com/ivas_rom_com.h

+2 −0

Original line number	Diff line number	Diff line
		@@ -159,9 +159,11 @@ extern const float nf_tw_smoothing_coeffs[N_LTP_GAIN_MEMS];
		* Stereo DTX ROM tables
		----------------------------------------------------------------------------------/

		extern const Word16 dft_cng_coh_pred_fx[][STEREO_DFT_COH_PRED_COEFFS];
		extern const float dft_cng_coh_pred[][STEREO_DFT_COH_PRED_COEFFS];
		extern const int16_t dft_cng_coh_u2i[9];
		extern const int16_t dft_cng_coh_i2u[9];
		extern const Word16 dft_cng_alpha_bits_fx[STEREO_DFT_N_COH_ALPHA_STEPS][STEREO_DFT_N_COH_ALPHA_LEVELS];
		extern const float dft_cng_alpha_bits[STEREO_DFT_N_COH_ALPHA_STEPS][STEREO_DFT_N_COH_ALPHA_LEVELS];
		extern const int16_t dft_cng_coh_alpha_start[STEREO_DFT_N_COH_ALPHA_STEPS - 1];