vad.h

AMR-NB 的编码实现,纯C, VC下建立工程即可用.

#ifndef VAD_H
#define VAD_H

#include "typedef.h"

#define VAD1                        /* VAD Algorithm Selection Compile Switch */


#define FRAME_LEN 160        /* Length (samples) of the input frame          */
#define COMPLEN 9               /* Number of sub-bands used by VAD              */
#define INV_COMPLEN 3641  /* 1.0/COMPLEN*2^15                             */
#define LOOKAHEAD 40         /* length of the lookahead used by speech coder */

#define UNITY 512                /* Scaling used with SNR calculation            */
#define UNIRSHFT 6              /* = log2(MAX_16/UNITY)                         */

#define TONE_THR (Word16)(0.65*MAX_16) /* Threshold for tone detection   */

/* Constants for background spectrum update */
#define ALPHA_UP1   (Word16)((1.0 - 0.95)*MAX_16)         /* Normal update, upwards:   */
#define ALPHA_DOWN1 (Word16)((1.0 - 0.936)*MAX_16)    /* Normal update, downwards  */
#define ALPHA_UP2   (Word16)((1.0 - 0.985)*MAX_16)        /* Forced update, upwards    */
#define ALPHA_DOWN2 (Word16)((1.0 - 0.943)*MAX_16)     /* Forced update, downwards  */
#define ALPHA3      (Word16)((1.0 - 0.95)*MAX_16)             /* Update downwards          */
#define ALPHA4      (Word16)((1.0 - 0.9)*MAX_16)               /* For stationary estimation */
#define ALPHA5      (Word16)((1.0 - 0.5)*MAX_16)               /* For stationary estimation */

/* Constants for VAD threshold */
#define VAD_THR_HIGH 1260   /* Highest threshold                 */
#define VAD_THR_LOW  720     /* Lowest threshold                  */
#define VAD_P1 0                    /* Noise level for highest threshold */
#define VAD_P2 6300               /* Noise level for lowest threshold  */
#define VAD_SLOPE (Word16)(MAX_16*(float)(VAD_THR_LOW-VAD_THR_HIGH)/(float)(VAD_P2-VAD_P1))

/* Parameters for background spectrum recovery function */
#define STAT_COUNT 20                   /* threshold of stationary detection counter         */
#define STAT_COUNT_BY_2 10         /* threshold of stationary detection counter         */
#define CAD_MIN_STAT_COUNT 5    /* threshold of stationary detection counter         */

#define STAT_THR_LEVEL 184    /* Threshold level for stationarity detection        */
#define STAT_THR 1000              /* Threshold for stationarity detection              */

/* Limits for background noise estimate */
#define NOISE_MIN 40              /* minimum */
#define NOISE_MAX 16000        /* maximum */
#define NOISE_INIT 150            /* initial */

/* Constants for VAD hangover addition */
#define HANG_NOISE_THR 100
#define BURST_LEN_HIGH_NOISE 4
#define HANG_LEN_HIGH_NOISE 7
#define BURST_LEN_LOW_NOISE 5
#define HANG_LEN_LOW_NOISE 4

/* Thresholds for signal power */
#define VAD_POW_LOW (Word32)15000         /* If input power is lower,              .   VAD is set to 0                         */
#define POW_PITCH_THR (Word32)343040     /* If input power is lower, pitch       .     detection is ignored                    */
#define POW_COMPLEX_THR (Word32)15000  /* If input power is lower, complex      . flags  value for previous frame  is un-set  */
 

/* Constants for the filter bank */
#define LEVEL_SHIFT 0       /* scaling                                  */
#define COEFF3   13363       /* coefficient for the 3rd order filter     */
#define COEFF5_1 21955     /* 1st coefficient the for 5th order filter */
#define COEFF5_2 6390       /* 2nd coefficient the for 5th order filter */

/* Constants for pitch detection */
#define LTHRESH 4
#define NTHRESH 4

/* Constants for complex signal VAD  */
#define CVAD_THRESH_ADAPT_HIGH  (Word16)(0.6 * MAX_16)  /* threshold for adapt stopping high    */
#define CVAD_THRESH_ADAPT_LOW  (Word16)(0.5 * MAX_16)   /* threshold for adapt stopping low     */
#define CVAD_THRESH_IN_NOISE  (Word16)(0.65 * MAX_16)     /* threshold going into speech on  a short term basis                   */

#define CVAD_THRESH_HANG  (Word16)(0.70 * MAX_16)      /* threshold                            */
#define CVAD_HANG_LIMIT  (Word16)(100)                          /* 2 second estimation time             */
#define CVAD_HANG_LENGTH  (Word16)(250)                       /* 5 second hangover                    */

#define CVAD_LOWPOW_RESET (Word16) (0.40 * MAX_16)     /* init in low power segment            */
#define CVAD_MIN_CORR (Word16) (0.40 * MAX_16)               /* lowest adaptation value              */

#define CVAD_BURST 20                                                                         /* speech burst length for speech reset */
#define CVAD_ADAPT_SLOW (Word16)(( 1.0 - 0.98) * MAX_16)              /* threshold for slow adaption */
#define CVAD_ADAPT_FAST (Word16)((1.0 - 0.92) * MAX_16)                /* threshold for fast adaption */
#define CVAD_ADAPT_REALLY_FAST (Word16)((1.0 - 0.80) * MAX_16)  /* threshold for really fastadaption                    */



/* state variable */
typedef struct {
   
   Word16 bckr_est[COMPLEN];     /* background noise estimate                */
   Word16 ave_level[COMPLEN];   /* averaged input components for stationary . estimation                            */
   Word16 old_level[COMPLEN];    /* input levels of the previous frame       */
   Word16 sub_level[COMPLEN];   /* input levels calculated at the end of   a frame (lookahead)                   */
   Word16 a_data5[3][2];             /* memory for the filter bank               */
   Word16 a_data3[5];                 /* memory for the filter bank               */

   Word16 burst_count;               /* counts length of a speech burst          */
   Word16 hang_count;               /* hangover counter                         */
   Word16 stat_count;                 /* stationary counter                       */

   /* Note that each of the following three variables (vadreg, pitch and tone)
      holds 15 flags. Each flag reserves 1 bit of the variable. The newest
      flag is in the bit 15 (assuming that LSB is bit 1 and MSB is bit 16). */
   Word16 vadreg;                    /* flags for intermediate VAD decisions     */
   Word16 pitch;                        /* flags for pitch detection                */
   Word16 tone;                         /* flags for tone detection                 */
   Word16 complex_high;         /* flags for complex detection              */
   Word16 complex_low;           /* flags for complex detection              */

   Word16 oldlag_count, oldlag;     /* variables for pitch detection            */
 
   Word16 complex_hang_count;   /* complex hangover counter, used by VAD    */
   Word16 complex_hang_timer;   /* hangover initiator, used by CAD          */
    
   Word16 best_corr_hp;                 /* FIP filtered value Q15                   */ 

   Word16 speech_vad_decision;   /* final decision                           */
   Word16 complex_warning;          /* complex background warning               */

   Word16 sp_burst_count;             /* counts length of a speech burst incl  HO addition                              */
   Word16 corr_hp_fast;                  /* filtered value                           */ 
} vadState1;


/***** Defines ****/

#define		YES		1
#define		NO		0
#define		ON		1
#define		OFF		0
#define		TRUE		1
#define		FALSE		0

#define         FRM_LEN                 80
#define         DELAY                   24
#define         FFT_LEN                 128

#define         NUM_CHAN                16
#define         LO_CHAN                 0
#define         HI_CHAN                 15

#define         UPDATE_THLD             35
#define         HYSTER_CNT_THLD         6
#define         UPDATE_CNT_THLD         50

#define		SHIFT_STATE_0		0		       /* channel energy scaled as 22,9 */
#define		SHIFT_STATE_1		1		       /* channel energy scaled as 27,4 */

#define		NOISE_FLOOR_CHAN_0	512		/* 1.0    scaled as 22,9 */
#define		MIN_CHAN_ENRG_0		32		/* 0.0625 scaled as 22,9 */
#define		MIN_NOISE_ENRG_0	32		       /* 0.0625 scaled as 22,9 */
#define		INE_NOISE_0		8192		/* 16.0   scaled as 22,9 */
#define		FRACTIONAL_BITS_0	9		       /* used as input to fn10Log10() */

#define		NOISE_FLOOR_CHAN_1	16		/* 1.0    scaled as 27,4 */
#define		MIN_CHAN_ENRG_1		1		/* 0.0625 scaled as 27,4 */
#define		MIN_NOISE_ENRG_1	1		       /* 0.0625 scaled as 27,4 */
#define		INE_NOISE_1		256		       /* 16.0   scaled as 27,4 */
#define		FRACTIONAL_BITS_1	4		       /* used as input to fn10Log10() */

#define		STATE_1_TO_0_SHIFT_R	(FRACTIONAL_BITS_1-FRACTIONAL_BITS_0)	/* state correction factor */
#define		STATE_0_TO_1_SHIFT_R	(FRACTIONAL_BITS_0-FRACTIONAL_BITS_1)	/* state correction factor */

#define         HIGH_ALPHA              29491		/* 0.9 scaled as 0,15 */
#define         LOW_ALPHA               22938		/* 0.7 scaled as 0,15 */
#define         ALPHA_RANGE             (HIGH_ALPHA - LOW_ALPHA)
#define         DEV_THLD                7168		       /* 28.0 scaled as 7,8 */

#define         PRE_EMP_FAC             (-26214)	       /* -0.8 scaled as 0,15 */

#define         CEE_SM_FAC              18022		/* 0.55 scaled as 0,15 */
#define         ONE_MINUS_CEE_SM_FAC    14746  /* 0.45 scaled as 0,15 */

#define         CNE_SM_FAC              3277		       /* 0.1 scaled as 0,15 */
#define         ONE_MINUS_CNE_SM_FAC    29491  /* 0.9 scaled as 0,15 */

#define         FFT_HEADROOM            2


typedef struct
{
	Word16 pre_emp_mem;
	Word16 update_cnt;
	Word16 hyster_cnt;
	Word16 last_update_cnt;
	Word16 ch_enrg_long_db[NUM_CHAN];	/* scaled as 7,8  */

	Word32 Lframe_cnt;
	Word32 Lch_enrg[NUM_CHAN];	/* scaled as 22,9 or 27,4 */
	Word32 Lch_noise[NUM_CHAN];	/* scaled as 22,9 */

	Word16 last_normb_shift;	      /* last block norm shift count */

	Word16 tsnr;			             /* total signal-to-noise ratio in dB (scaled as 7,8) */
	Word16 hangover;
	Word16 burstcount;
	Word16 fupdate_flag;		/* forced update flag from previous frame */
	Word16 negSNRvar;		/* Negative SNR variance (scaled as 7,8) */
	Word16 negSNRbias;		/* sensitivity bias from negative SNR variance (scaled as 15,0) */

	Word16 shift_state;		/* use 22,9 or 27,4 scaling for ch_enrg[] */

	Word32 L_R0;
	Word32 L_Rmax;
	Flag   LTP_flag;		      /* Use to indicate the the LTP gain is > LTP_THRESH */

} vadState2;


/*********************************************************************************
*                         DECLARATION OF PROTOTYPES
*********************************************************************************/
int vad1_init (vadState1 **st); 
int vad1_reset (vadState1 *st);
void vad1_exit (vadState1 **st);
Word16 vad1 (vadState1 *st,   Word16 in_buf[]  );
Word16	vad2 (Word16 *farray_ptr, vadState2 *st);
int	vad2_init (vadState2 **st);
void	vad2_exit (vadState2 **state);

void	r_fft (Word16 *farray_ptr);

#ifndef	VAD2
#define vadState vadState1
#else 
#define vadState vadState2

#endif
#endif