Merge branch 'generate_stereo_masking_noise_fxd' into 'main' (e7a2525e) · Commits · SA4 / Audio / IVAS BASOP

lib_com/cnst.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -1666,6 +1666,7 @@ enum
		#define FD_CNG_JOINT_stages_25bits 4

		#define OUTMAX_INV 0.000030517578125f /* 1/2^15 */
		#define OUTMAX_INV_FX 65536 /* 1/2^15 (Q31) */
		#define OUTMAX_SQ 1073741824.f /* 2^30 */
		#define OUTMAX_SQ_INV 0.00000000093132257461547852f /* 1/2^30 */

lib_com/fd_cng_com.c

+130 −2

Original line number	Diff line number	Diff line
		@@ -39,10 +39,14 @@
		#include "options.h"
		#include <math.h>
		#include "prot.h"
		#include "prot_fx2.h"
		#include "rom_com.h"
		#include "wmc_auto.h"


		#ifdef IVAS_FLOAT_FIXED
		#define FFT_SCALING_512 1073741824 //Q22
		#define FFT_SCALING_640 1342177280 //Q22
		#endif
		/*-------------------------------------------------------------------
		* Local function prototypes
		-------------------------------------------------------------------/
		@@ -103,6 +107,10 @@ void initFdCngCom_flt(
		set_f( hFdCngCom->sidNoiseEst_flt, 0.0f, NPART );
		set_f( hFdCngCom->A_cng_flt, 0.0f, M + 1 );
		hFdCngCom->A_cng_flt[0] = 1.f;
		#ifdef IVAS_FLOAT_FIXED
		set_s( hFdCngCom->A_cng, 0, M + 1 );
		hFdCngCom->A_cng[0] = MAX_16;
		#endif

		/* Set some counters and flags */
		hFdCngCom->inactive_frame_counter = 0; /* Either SID or zero frames */
		@@ -110,6 +118,9 @@ void initFdCngCom_flt(
		hFdCngCom->frame_type_previous = ACTIVE_FRAME;
		hFdCngCom->flag_noisy_speech = 0;
		hFdCngCom->likelihood_noisy_speech_flt = 0.f;
		#ifdef IVAS_FLOAT_FIXED
		hFdCngCom->likelihood_noisy_speech_32fx = 0;
		#endif
		hFdCngCom->numCoreBands = 0;
		hFdCngCom->stopBand = 0;
		hFdCngCom->startBand = 0;
		@@ -937,6 +948,115 @@ void SynthesisSTFT_flt(
		return;
		}

		#ifdef IVAS_FLOAT_FIXED
		/*-------------------------------------------------------------------
		* SynthesisSTFT_fx()
		*
		* STFT synthesis filterbank
		-------------------------------------------------------------------/

		void SynthesisSTFT_fx(
		Word32 fftBuffer, / i : FFT bins */
		Word16 Q_in,
		Word32 *timeDomainOutput,
		Word32 *olapBuffer,
		const Word16 *olapWin,
		const int16_t tcx_transition,
		HANDLE_FD_CNG_COM hFdCngCom, /* i/o: FD_CNG structure containing all buffers and variables */
		const int16_t element_mode, /* i : element mode */
		const int16_t nchan_out /* i : number of output channels */
		)
		{
		int16_t i;
		Word32 buf_fx[M + 1 + 320], tmp_fx;

		/* Perform IFFT */
		RFFTN_fx( fftBuffer, hFdCngCom->fftSineTab_fx, hFdCngCom->fftlen, 1 );

		/* Handle overlap in P/S domain for stereo */
		IF( ( element_mode == IVAS_CPE_TD \|\| element_mode == IVAS_CPE_DFT ) && nchan_out == 2 )
		{
		mvl2l( olapBuffer + 3 * hFdCngCom->frameSize / 4 - ( M + 1 ), buf_fx, hFdCngCom->frameSize + M + 1 );
		set_l( olapBuffer, 0, hFdCngCom->fftlen );
		}
		ELSE
		{
		mvl2l( olapBuffer + hFdCngCom->frameSize, olapBuffer, hFdCngCom->frameSize );
		set_l( olapBuffer + hFdCngCom->frameSize, 0, hFdCngCom->frameSize ); /olapBuffer, fftBuffer, olapWin/
		}

		IF( tcx_transition )
		{
		FOR( i = 0; i < 5 * hFdCngCom->frameSize / 4; i++ )
		{
		olapBuffer[i] = fftBuffer[i];
		}
		}
		ELSE
		{
		FOR( i = hFdCngCom->frameSize / 4; i < 3 * hFdCngCom->frameSize / 4; i++ )
		{
		olapBuffer[i] = L_add( olapBuffer[i], Mpy_32_16_1( fftBuffer[i], olapWin[i - hFdCngCom->frameSize / 4] ) );
		}
		FOR( ; i < 5 * hFdCngCom->frameSize / 4; i++ )
		{
		olapBuffer[i] = fftBuffer[i];
		}
		}
		FOR( ; i < 7 * hFdCngCom->frameSize / 4; i++ )
		{
		olapBuffer[i] = Mpy_32_16_1( fftBuffer[i], olapWin[i - 3 * hFdCngCom->frameSize / 4] );
		}

		FOR( ; i < hFdCngCom->fftlen; i++ )
		{
		olapBuffer[i] = 0;
		}

		Word32 fftScale = 0;
		SWITCH( hFdCngCom->fftlen )
		{
		case 640:
		fftScale = FFT_SCALING_640;
		break;
		case 512:
		fftScale = FFT_SCALING_512;
		break;
		default:
		assert( !"Not supported FFT length!" );
		}
		/* Get time-domain signal */
		// v_multc(olapBuffer + hFdCngCom->frameSize / 4, (float)(hFdCngCom->fftlen / 2), timeDomainOutput, hFdCngCom->frameSize);
		v_multc_fixed( olapBuffer + hFdCngCom->frameSize / 4, fftScale, timeDomainOutput, hFdCngCom->frameSize ); // Q_in - 9
		/* Get excitation */
		IF( ( element_mode == IVAS_CPE_TD \|\| element_mode == IVAS_CPE_DFT ) && nchan_out == 2 )
		{
		FOR( i = 0; i < hFdCngCom->frameSize / 2; i++ )
		{
		buf_fx[i + ( M + 1 )] = L_add( buf_fx[i + ( M + 1 )], olapBuffer[i + hFdCngCom->frameSize / 4] );
		}
		// v_multc(buf, (float)(hFdCngCom->fftlen / 2), buf, M + 1 + hFdCngCom->frameSize);
		v_multc_fixed( buf_fx, fftScale, buf_fx, M + 1 + hFdCngCom->frameSize );
		}
		ELSE
		{
		// v_multc(olapBuffer + hFdCngCom->frameSize / 4 - (M + 1), (float)(hFdCngCom->fftlen / 2), buf, M + 1 + hFdCngCom->frameSize);
		v_multc_fixed( olapBuffer + ( hFdCngCom->frameSize / 4 ) - ( M + 1 ), fftScale, buf_fx, M + 1 + hFdCngCom->frameSize );
		}

		tmp_fx = buf_fx[0];
		// preemph(buf + 1, PREEMPH_FAC_FLT, M + hFdCngCom->frameSize, &tmp);
		preemph_ivas_fx( buf_fx + 1, PREEMPH_FAC, M + hFdCngCom->frameSize, &tmp_fx );
		// residu(hFdCngCom->A_cng_flt, M, buf + 1 + M, hFdCngCom->exc_cng_flt, hFdCngCom->frameSize);
		residu_ivas_fx( hFdCngCom->A_cng, Q13, M, buf_fx + 1 + M, hFdCngCom->exc_cng_32fx, hFdCngCom->frameSize );
		for ( i = 0; i < hFdCngCom->frameSize; i++ )
		{
		hFdCngCom->exc_cng_flt[i] = fix_to_float( hFdCngCom->exc_cng_32fx[i], Q_in - 9 );
		}

		return;
		}
		#endif

		/*-------------------------------------------------------------------
		* SynthesisSTFT_dirac_flt()
		@@ -1112,6 +1232,9 @@ void lpc_from_spectrum_flt(
		int16_t fftlen = hFdCngCom->fftlen;
		const float *fftSineTab = hFdCngCom->fftSineTab_flt;
		float *A = hFdCngCom->A_cng_flt;
		#ifdef IVAS_FLOAT_FIXED
		Word16 *A_fx = hFdCngCom->A_cng;
		#endif

		/* Power Spectrum */
		ptr = fftBuffer;
		@@ -1165,7 +1288,12 @@ void lpc_from_spectrum_flt(

		/* LPC */
		lev_dur( A, r, M, NULL );

		#ifdef IVAS_FLOAT_FIXED
		for ( i = 0; i < M + 1; i++ )
		{
		A_fx[i] = float_to_fix16( A[i], Q13 );
		}
		#endif
		return;
		}

lib_com/ivas_cnst.h

+2 −0

Original line number	Diff line number	Diff line
		@@ -50,6 +50,8 @@
		#define _180_OVER_PI_Q25 1922527233
		#define PI_OVER_4_Q29 421657440
		#define PI_OVER_Q29 1686629760
		#define Q31_0_99 2126008811
		#define Q31_0_01 21474836
		#endif

		#define SQRT2 1.414213562373095f

lib_com/preemph.c

+29 −0

Original line number	Diff line number	Diff line
		@@ -66,3 +66,32 @@ void preemph(

		return;
		}


		/-------------------------------------------------------------
		* preemph_ivas_fx()
		*
		* Preemphasis: filtering through 1 - mu z^-1
		-------------------------------------------------------------/

		void preemph_ivas_fx(
		Word32 signal, / i/o: signal */
		const Word16 mu, /* i : preemphasis factor */
		const Word16 L, /* i : vector size */
		Word32 mem / i/o: memory (x[-1]) */
		)
		{
		Word16 i;
		Word32 temp;

		temp = signal[L - 1];
		FOR( i = L - 1; i > 0; i-- )
		{
		signal[i] = L_sub( signal[i], Mpy_32_16_1( signal[i - 1], mu ) );
		}

		signal[0] = L_sub( signal[0], Mpy_32_16_1( *mem, mu ) );
		*mem = temp;

		return;
		}
		No newline at end of file

lib_com/prot.h

+56 −1

Original line number	Diff line number	Diff line
		@@ -813,7 +813,12 @@ void preemph(
		const int16_t L, /* i : vector size */
		float mem / i/o: memory (x[-1]) */
		);

		void preemph_ivas_fx(
		Word32 signal, / i/o: signal */
		const Word16 mu, /* i : preemphasis factor */
		const Word16 L, /* i : vector size */
		Word32 mem / i/o: memory (x[-1]) */
		);
		void cb_shape(
		const int16_t preemphFlag, /* i : flag for pre-emphasis */
		const int16_t pitchFlag, /* i : flag for pitch sharpening */
		@@ -1102,6 +1107,15 @@ void residu(
		const int16_t l /* i : size of filtering */
		);

		void residu_ivas_fx(
		const Word16 a, / i : LP filter coefficients */
		const Word16 a_exp,
		const Word16 m, /* i : order of LP filter */
		const Word32 x, / i : input signal (usually speech) */
		Word32 y, / o : output signal (usually residual) */
		const Word16 l /* i : size of filtering */
		);

		void calc_residu(
		const float speech, / i : weighted speech signal */
		float res, / o : residual signal */
		@@ -8635,6 +8649,21 @@ void generate_masking_noise_flt(
		const int16_t nchan_out /* i : number of output channels */
		);

		#ifdef IVAS_FLOAT_FIXED
		void generate_masking_noise_ivas_fx(
		Word32 timeDomainBuffer, / i/o: time-domain signal */
		Word16 exp_out, / o : time-domain signal exp */
		HANDLE_FD_CNG_COM hFdCngCom, /* i/o: FD_CNG structure containing all buffers and variables */
		const int16_t length, /* i : frame size */
		const int16_t core, /* i : core */
		const int16_t return_noise, /* i : noise is returned instead of added */
		const int16_t secondary, /* i : flag to indicate secondary noise generation */
		const int16_t element_mode, /* i : element mode */
		STEREO_CNG_DEC_HANDLE hStereoCng, /* i : stereo CNG handle */
		const int16_t nchan_out /* i : number of output channels */
		);
		#endif

		void generate_masking_noise_update_seed_flt(
		HANDLE_FD_CNG_COM hFdCngCom /* i/o: FD_CNG structure containing all buffers and variables */
		);
		@@ -8674,6 +8703,18 @@ void generate_stereo_masking_noise(
		const int16_t nchan_out /* i : number of output channels */
		);

		#ifdef IVAS_FLOAT_FIXED
		void generate_stereo_masking_noise_fx(
		float syn, / i/o: time-domain signal */
		Decoder_State st, / i/o: decoder state structure */
		STEREO_TD_DEC_DATA_HANDLE hStereoTD, /* i : TD stereo structure */
		const int16_t flag_sec_CNA, /* i : CNA flag for secondary channel */
		const int16_t fadeOut, /* i : only fade out of previous state */
		STEREO_CNG_DEC_HANDLE hStereoCng, /* i : stereo CNG handle */
		const int16_t nchan_out /* i : number of output channels */
		);
		#endif

		void apply_scale_flt(
		float scale, / i : scale factor */
		const int16_t bwidth, /* i : audio bandwidth */
		@@ -8728,6 +8769,20 @@ void SynthesisSTFT_flt(
		const int16_t nchan_out /* i : number of output channels */
		);

		#ifdef IVAS_FLOAT_FIXED
		void SynthesisSTFT_fx(
		Word32 fftBuffer, / i : FFT bins */
		Word16 Q_in,
		Word32 *timeDomainOutput,
		Word32 *olapBuffer,
		const Word16 *olapWin,
		const int16_t tcx_transition,
		HANDLE_FD_CNG_COM hFdCngCom, /* i/o: FD_CNG structure containing all buffers and variables */
		const int16_t element_mode, /* i : element mode */
		const int16_t nchan_out /* i : number of output channels */
		);
		#endif

		float rand_gauss_flt(
		float *x,
		int16_t *seed );