/*====================================================================================
EVS Codec 3GPP TS26.443 Jun 30, 2015. Version CR 26.443-0006
====================================================================================*/
#include <math.h>
#include "options.h"
#include "cnst.h"
#include "prot.h"
#include "rom_enc.h"
/*-----------------------------------------------------------------*
* Local constants
*-----------------------------------------------------------------*/
#define HANGOVER_LONG 10 /* Hangover for CNG */
#define HANGOVER_LONG_HE 20 /* Hangover of CNG */
#define HANGOVER_LONG_MUSIC 20 /* Hangover of CNG */
#define HANGOVER_LONG_NB 8 /* Hangover for CNG */
#define ACTIVE_FRAMES 3 /* Number of consecutive active SPEECH frames necessary to trigger HO */
#define TH16_2 35.0f /* long-term SNR that separates the curves for clean speech and noisy speech */
#define TH8_1 20.0f /* long-term SNR that separates the curves for clean speech and noisy speech */
#define TH16_2_NFLAG 35.0f
#define TH8_1_NFLAG 35.0f
#define SNR_OUTLIER_WGHT_1 1.00f
#define SNR_OUTLIER_WGHT_2 1.01f
#define SNR_OUTLIER_WGHT_3 1.02f
#define OUTLIER_THR_1 10.0f
#define OUTLIER_THR_2 6.0f
#define MAX_SNR_OUTLIER_IND 17
#define MAX_SNR_OUTLIER_1 10.0f
#define MAX_SNR_OUTLIER_2 25.0f
#define MAX_SNR_OUTLIER_3 50.0f
/*---------------------------------------------------------------------*
* wb_vad_init()
*
* VAD initializations
*---------------------------------------------------------------------*/
void wb_vad_init(
short *nb_active_frames, /* o : nb of consecutive active speech frames */
short *hangover_cnt,
float *lp_speech, /* o : long-term active speech level */
short *nb_active_frames_he, /* o : nb of consecutive active speech frames */
short *hangover_cnt_he,
float *bcg_flux, /* o : background noise fluctuation */
short *soft_hangover, /* o : soft hangover counter */
short *voiced_burst, /* o : consecutive voiced speech counter */
short *bcg_flux_init, /* o : initialization period for noise fluctuation estimation */
short *nb_active_frames_he1, /* o : nb of consecutive active speech frames 1 */
short *hangover_cnt_he1,
long *vad_flag_reg_H,
long *vad_flag_reg_L,
long *vad_prim_reg,
short *vad_flag_cnt_50,
short *vad_prim_cnt_16,
short *hangover_cnt_dtx,
short *flag_noisy_speech_snr
,short *hangover_cnt_music /* o : counter of VAD DTX Music hangover frames */
)
{
*hangover_cnt = 0; /* Hangover counter initialized to 0 */
*nb_active_frames = ACTIVE_FRAMES; /* The counter of SPEECH frames necessary to trigger HO */
/* is set to max (-> start with hangover) */
*lp_speech = 45.0f; /* Initialize the long-term active speech level in dB */
*flag_noisy_speech_snr=0;
*vad_flag_reg_H = 0L;
*vad_flag_reg_L = 0L;
*vad_prim_reg = 0L;
*vad_flag_cnt_50 = 0;
*vad_prim_cnt_16 = 0;
/* By default one should not start with a hangover */
*hangover_cnt_dtx = HANGOVER_LONG; /* hangover for DTX */
*hangover_cnt_music = HANGOVER_LONG_MUSIC; /* hangover for DTX */
*hangover_cnt_he = 0; /* Hangover counter initialized to 0 */
*nb_active_frames_he = ACTIVE_FRAMES; /* The counter of SPEECH frames necessary to trigger HO */
*bcg_flux = 70;
*soft_hangover = 0;
*voiced_burst = 0;
*bcg_flux_init = 50;
*nb_active_frames_he1 = ACTIVE_FRAMES;
*hangover_cnt_he1 = 0;
return;
}
/*-----------------------------------------------------------------*
* sing_thr_snr_acc()
*
* accumulate snr_sum with significance thresholds
*-----------------------------------------------------------------*/
static void sign_thr_snr_acc(
float *snr_sum,
float snr,
float sign_thr,
float min_snr
)
{
if( snr >= sign_thr )
{
*snr_sum = *snr_sum + snr;
}
else
{
*snr_sum = *snr_sum + min_snr;
}
return;
}
/*-----------------------------------------------------------------*
* dtx_hangover_addition()
*
* accumulate snr_sum with significance thresholds
*-----------------------------------------------------------------*/
short dtx_hangover_addition(
Encoder_State *st, /* i/o: encoder state structure */
const short localVAD,
const short vad_flag,
const float lp_snr,
const short cldfb_subtraction,
short *vad_hover_flag
)
{
short hangover_short_dtx, flag_dtx;
short ho_limit_clean;
flag_dtx = 0;
/* Determine initial hangover length */
hangover_short_dtx = 2; /* was 1 */
if ( ( lp_snr < 16.0f && st->input_bwidth != NB ) ||
st->prim_act_he > 0.95f )
{
hangover_short_dtx = 3; /* was 2 */
}
/* Adjust hangover according to activity history */
if (st->vad_prim_cnt_16 > 12 ) /* 12 requires roughly > 80% primary activity */
{
hangover_short_dtx = hangover_short_dtx + 2;
}
if (st->vad_flag_cnt_50 >40 ) /* 40 requires roughtly > 80% flag activity */
{
hangover_short_dtx = hangover_short_dtx + 5;
}
/* Keep hangover_short lower than maximum hangover count */
if (hangover_short_dtx > HANGOVER_LONG-1)
{
hangover_short_dtx = HANGOVER_LONG-1;
}
/* Only allow short HO if not sufficient active frames */
ho_limit_clean = 3;
if (st->core == AMR_WB_CORE )
{
ho_limit_clean = 2;
}
if ( st->input_bwidth != NB && st->core != AMR_WB_CORE && lp_snr > 25.0f )
{
ho_limit_clean = 2;
}
if ( ho_limit_clean != 0 )
{
if ( (hangover_short_dtx > ho_limit_clean) && ( ( st->vad_prim_cnt_16 < 7 ) || ( lp_snr > 16 && st->prim_act_he < 0.85 ) ) )
{
hangover_short_dtx = ho_limit_clean;
}
}
/* hangover adjustment from combined FFT + CLDFBVAD */
if (st->core != AMR_WB_CORE)
{
hangover_short_dtx = hangover_short_dtx - cldfb_subtraction;
if( hangover_short_dtx < 0 )
{
hangover_short_dtx = 0;
}
}
if ( vad_flag == 1 ) /* Speech present */
{
flag_dtx = 1;
/* Add hangover after sufficient # of active frames or sufficient activity during last second */
if (st->nb_active_frames >= ACTIVE_FRAMES || st->vad_flag_cnt_50 >45 ) /* 45 requires roughtly > 90% flag activity */
{
st->hangover_cnt_dtx = 0;
}
/* inside HO period */
if( st->hangover_cnt_dtx < HANGOVER_LONG && st->hangover_cnt_dtx != 0 )
{
st->hangover_cnt_dtx++;
}
st->hangover_terminate_flag = 0;
/* Music hangover when music detected */
if( st->prim_act_he > 0.98f && st->Etot_lp > 40 && st->vad_prim_cnt_16 > 14 && st->vad_flag_cnt_50 > 48 )
{
st->hangover_cnt_music = 0;
}
/* inside music HO period */
if (st->hangover_cnt_music < HANGOVER_LONG_MUSIC && st->hangover_cnt_music != 0 )
{
st->hangover_cnt_music++;
}
}
else
{
/* Reset the counter of speech frames necessary to start hangover algorithm */
if(st->hangover_cnt_dtx < HANGOVER_LONG ) /* inside HO period */
{
st->hangover_cnt_dtx++;
}
if(st->hangover_cnt_music < HANGOVER_LONG_MUSIC ) /* inside music HO period */
{
st->hangover_cnt_music++;
}
/* fast terminate DTX hangover if st->hangover_terminate_flag is set */
if ( st->hangover_terminate_flag == 1 )
{
st->hangover_cnt = HANGOVER_LONG;
st->hangover_cnt_dtx = HANGOVER_LONG;
st->hangover_terminate_flag = 0;
/* only shorten music hangover when low energy frames */
if ( st->Etot_lp < 20.0f )
{
st->hangover_cnt_music = HANGOVER_LONG_MUSIC;
}
}
if( st->hangover_cnt_dtx <= hangover_short_dtx ) /* "hard" hangover */
{
flag_dtx = 1;
}
if( st->hangover_cnt_music <= 15 ) /* "hard" hangover */
{
flag_dtx = 1;
}
}
if ( flag_dtx != 0 && localVAD == 0 )
{
*vad_hover_flag = 1;
}
return flag_dtx ;
}
/*-----------------------------------------------------------------*
* wb_vad()
*
* Voice Activity Detector
*-----------------------------------------------------------------*/
short wb_vad(
Encoder_State *st, /* i/o: encoder state structure */
const float fr_bands[], /* i : per band input energy (contains 2 vectors) */
short *localVAD,
short *noisy_speech_HO, /* o : SC-VBR noisy speech HO flag */
short *clean_speech_HO, /* o : SC-VBR clean speech HO flag */
short *NB_speech_HO, /* o : SC-VBR NB speech HO flag */
float *snr_sum_he, /* o : Output snr_sum as weighted spectral measure */
short *localVAD_HE_SAD, /* o : HE_SAD decision without hangovers */
short *flag_noisy_speech_snr /* o: */
)
{
short i, j, flag, hangover_short;
float snr[NB_BANDS], snr_sum, thr1, thr2, lp_snr, nk, nc, th_clean;
const float *pt1, *pt2, *pt3;
float min_snr, sign_thr;
float fr_enr;
float ftmp, ftmp1;
float mssnr = 0;
float snr_sumt;
float vad_thr;
short hangover_hd;
short snr_idx;
float delta1, delta2, delta3;
short flag_he1;
float mssnr_hov;
short stmp;
float msnr;
float snr_outlier;
short snr_outlier_index;
float accum_ener_L, accum_ener_H;
float delta4;
float snr18=1.0f, snr19=1.0f;
short nb_sig_snr;
float nv;
float snr_sum_HE_SAD;
float sign_thr_HE_SAD,min_snr_HE_SAD;
float nv_ofs;
float thr1_ol;
float snr_sum_ol;
snr_outlier = 0;
snr_outlier_index = 0;
accum_ener_L = 0;
accum_ener_H = 0;
if( st->input_bwidth == NB )
{
st->min_band = 1;
st->max_band = 16;
}
else
{
st->min_band = 0;
st->max_band = 19;
}
/*---------------------------------------------------------------------*
* set SNR thresholds depending on the input rate
*---------------------------------------------------------------------*/
if(st->max_band == 19 ) /* WB input */
{
nk = 0.1f;
nc = 16.1f;
nv = 2.05f;
nv_ofs = 1.65f;
th_clean = TH16_2;
if ( st->input_bwidth == WB )
{
sign_thr = 1.3f;
min_snr = 0.8f;
}
else
{
sign_thr = 1.75f;
min_snr = 0.25f;
}
sign_thr_HE_SAD = 2.5f;
min_snr_HE_SAD = 0.2f;
}
else /* NB input */
{
nk = 0.10f;
nc = 16.0f;
nv = 4.00f; /* Was 4.5f but trunkated to 4.00 used when converted to short */
nv_ofs = 1.15f;
th_clean = TH8_1;
sign_thr = 1.75f;
min_snr = 0.25f;
sign_thr_HE_SAD = 2.65f;
min_snr_HE_SAD = 0.05f;
}
hangover_short = 0;
if( st->Opt_SC_VBR )
{
*noisy_speech_HO = 0;
*clean_speech_HO = 0;
*NB_speech_HO = 0;
}
/*---------------------------------------------------------------------*
* compute SNR for each band & total
*---------------------------------------------------------------------*/
pt1 = fr_bands;
pt2 = fr_bands + NB_BANDS;
snr_sum = 0.0f;
*snr_sum_he = 0.0f;
snr_sumt = 0;
mssnr_hov = 0;
snr_sum_HE_SAD = 0.0f;
lp_snr = st->lp_speech - st->lp_noise;
if ( lp_snr > 24.0f )
{
snr_idx = 0;
}
else if ( lp_snr > 18 )
{
snr_idx = 1;
}
else
{
snr_idx = 2;
}
if ( snr_idx == 0 )
{
stmp = 6;
delta1 = 0.0f;
delta2 = 0.0f;
delta3 = 0.0f;
delta4 = 0.0f;
vad_thr = 2.4f*lp_snr - 42.2f;
vad_thr = min(vad_thr, 80);
}
else if ( snr_idx == 1)
{
stmp = 6;
delta1 = 0.1f;
delta2 = 0.2f;
delta3 = 0.2f;
delta4 = 0.2f;
vad_thr = 2.4f*lp_snr - 40.2f;
vad_thr = min(vad_thr, 80);
}
else
{
stmp = 9;
delta1 = 0.2f;
delta2 = 0.4f;
delta3 = 0.3f;
delta4 = 0.4f;
vad_thr = 2.5f*lp_snr - 10.0f;
vad_thr = max(vad_thr, 1);
}
pt3 = st->bckr;
nb_sig_snr = 20;
for( i=st->min_band; i<=st->max_band; i++ )
{
ftmp = *pt1++;
ftmp1 = *pt2++;
fr_enr = ( 0.2f * st->enrO[i] + 0.4f * ftmp + 0.4f * ftmp1 );
if (ftmp > ftmp1)
{
snr[i] = ( 0.2f * st->enrO[i] + 0.4f * ftmp + 0.4f * ftmp1 ) / *pt3++;
}
else
{
snr[i] = ( 0.2f * st->enrO[i] + 0.3f * ftmp + 0.5f * ftmp1 ) / *pt3++;
}
if ( snr[i] < 2.0f )
{
nb_sig_snr--;
}
if ( snr[i] < 1 )
{
snr[i] = 1;
}
snr[i] = (float)log10(snr[i]);
snr_sumt += snr[i];
if (i < 2)
{
ftmp = snr[i] + delta1;
}
else if (i < 7)
{
ftmp = snr[i] + delta2;
}
else if (i <18)
{
ftmp = snr[i] + delta3;
}
else
{
ftmp = snr[i] + delta4;
}
ftmp1 = ftmp;
if ( i < 7 )
{
ftmp1 = ftmp + 0.4f;
}
ftmp = min(ftmp, 2.0f);
ftmp1 = min(ftmp1, 2.0f);
msnr = 1;
for (j=0; j<stmp; j++)
{
msnr *= ftmp;
}
mssnr += msnr;
if ( i == 18 )
{
snr18 = msnr;
}
else if ( i == 19 )
{
snr19 = msnr;
}
msnr = 1;
for (j=0; j<stmp; j++)
{
msnr *= ftmp1;
}
mssnr_hov += msnr;
snr[i] = fr_enr / st->bckr[i];
sign_thr_snr_acc( &snr_sum_HE_SAD, snr[i], sign_thr_HE_SAD, min_snr_HE_SAD );
sign_thr_snr_acc( &snr_sum,snr[i], sign_thr, min_snr );
/* To make snr[] compatible with older versions where snr[i] >= 1
also this could be removed if this no longer is a requriement */
if( snr[i] < 1.0f )
{
snr[i] = 1.0f;
}
/* accumulate background noise energy in bands [0-2] and in bands [3-19]*/
if( i < 3 )
{
accum_ener_L = accum_ener_L + st->bckr[i];
}
else
{
accum_ener_H = accum_ener_H + st->bckr[i];
}
/* identify the outlier band */
if( snr[i] > snr_outlier )
{
snr_outlier = snr[i];
snr_outlier_index = i;
}
}
if ( (st->max_band == 19) && (snr[18] > 5.0f) && (snr[19] > 5.0f) )
{
ftmp = (mssnr + 3*(snr18 + snr19)) * 0.77f;
if ( ftmp > mssnr )
{
mssnr = ftmp;
}
}
else if ( snr_idx != 0 && nb_sig_snr > 13 )
{
if ( 2.5f*lp_snr - 15.5f > 0 )
{
mssnr += 2.5f*lp_snr - 15.5f;
}
}
/* Separate SNR_SUM modification to */
snr_sum_ol = snr_sum;
if(st->max_band == 19 && snr_outlier < MAX_SNR_OUTLIER_3 && snr_outlier_index > 3 && snr_outlier_index < MAX_SNR_OUTLIER_IND) /* Update the total SNR only for WB signals */
{
if( (accum_ener_L > OUTLIER_THR_1 * accum_ener_H ) || (snr_outlier < MAX_SNR_OUTLIER_1) )
{
snr_sum_ol = SNR_OUTLIER_WGHT_1 * (snr_sum_ol - snr_outlier);
}
else if( (accum_ener_L > OUTLIER_THR_2 * accum_ener_H ) || (snr_outlier < MAX_SNR_OUTLIER_2) )
{
snr_sum_ol = SNR_OUTLIER_WGHT_2 * (snr_sum_ol - snr_outlier);
}
else
{
snr_sum_ol = SNR_OUTLIER_WGHT_3 * (snr_sum_ol - snr_outlier);
}
}
st->snr_sum_vad = 0.5f * st->snr_sum_vad + 0.5f * snr_sum_ol;
snr_sum_ol = 10.0f * (float)log10( snr_sum_ol );
snr_sum = snr_sum_ol; /* for NB no outlier modification */
snr_sum_HE_SAD = 10.0f * (float)log10( snr_sum_HE_SAD );
*snr_sum_he=snr_sum_HE_SAD;
/*---------------------------------------------------------------------*
* compute thr1 for SAD decision
*---------------------------------------------------------------------*/
lp_snr = st->lp_speech - st->lp_noise; /* long-term SNR */
if (lp_snr < st->sign_dyn_lp)
{
lp_snr +=1;
if (lp_snr > st->sign_dyn_lp)
{
lp_snr = st->sign_dyn_lp;
}
}
thr1 = nk * lp_snr + nc + nv * ( st->Etot_v_h2 - nv_ofs); /* Linear function for noisy speech */
if (lp_snr > 20.0f )
{
thr1 = thr1 + 0.3f * (lp_snr - 20.0f);
if ( st->max_band==16 && lp_snr > 40 && thr1 > 24.1f && st->lp_speech < 45.0f )
{
thr1 = 24.1f;
}
}
/*---------------------------------------------------------------------*
* WB input
* SNR threshold computing
* Hangover control & final VAD decision
*---------------------------------------------------------------------*/
if( st->input_bwidth != NB )
{
/* Outlier Detection first calculates thr1_ol and snr_sum_ol instead of
thr1 and snr_sum */
thr1_ol = thr1;
if( lp_snr < th_clean )
{
hangover_short = 4;
if( ( snr_outlier_index <= 4 && (st->last_coder_type > UNVOICED) && !st->Opt_SC_VBR ) ||
( snr_outlier_index <= 4 && (st->last_7k2_coder_type > UNVOICED) && st->Opt_SC_VBR ) )
{
thr1_ol = thr1 - 1.0f ;
snr_sum_ol = 10.0f * (float)log10( st->snr_sum_vad );
}
else if ( ( (st->last_coder_type <= UNVOICED) && (snr_outlier < MAX_SNR_OUTLIER_2) && !st->Opt_SC_VBR ) ||
( (st->last_7k2_coder_type <= UNVOICED) && (snr_outlier < MAX_SNR_OUTLIER_2) && st->Opt_SC_VBR ) )
{
thr1_ol = thr1 + (float)(1.0f - 0.04f * snr_outlier);
}
else
{
thr1_ol = thr1 + max(0, (float)(0.6f - 0.01f * snr_outlier));
}
}
else
{
if( st->Opt_SC_VBR )
{
hangover_short = 3;
}
else
{
hangover_short = 3;
}
}
/* The use of outlier detection had been removed by accident at some point */
snr_sum = snr_sum_ol;
thr1 = thr1_ol;
/* DTX HANGOVER ADDITION MOVED TO pre_proc() */
flag_he1 = 0;
*localVAD = 0;
if ( mssnr > vad_thr )
{
*localVAD = 1; /* he1 primary decision */
flag_he1 = 1;
st->nb_active_frames_he1++; /* Counter of consecutive active speech frames */
if ( st->nb_active_frames_he1 >= ACTIVE_FRAMES )
{
st->nb_active_frames_he1 = ACTIVE_FRAMES;
st->hangover_cnt_he1 = 0; /* Reset the counter of hangover frames after at least "active_frames" speech frames */
}
/* inside HO period */
if ( st->hangover_cnt_he1 < HANGOVER_LONG_HE && st->hangover_cnt_he1 != 0 )
{
st->hangover_cnt_he1++;
}
if ( st->soft_hangover > 0 )
{
st->soft_hangover--;
}
}
else
{
/* Reset the counter of speech frames necessary to start hangover algorithm */
st->nb_active_frames_he1 = 0;
}
if ( st->voiced_burst > 3 )
{
if ( st->bcg_flux < 40 )
{
st->soft_hangover = hangover_sf_tbl[snr_idx+3];
}
else
{
st->soft_hangover = hangover_sf_tbl[snr_idx];
}
}
hangover_hd = hangover_hd_tbl[snr_idx];
if ( st->bcg_flux < 40 )
{
hangover_hd = (hangover_hd>>1) + 1;
}
if ( flag_he1 == 0 && st->soft_hangover > 0 )
{
if ( mssnr_hov > vad_thr )
{
flag_he1 = 1;
st->soft_hangover--;
}
else
{
st->soft_hangover = 0;
}
if ( st->soft_hangover < 0 )
{
st->soft_hangover = 0;
}
}
if ( flag_he1 == 0 && st->hangover_cnt_he1 < hangover_hd && st->soft_hangover == 0 )
{
flag_he1 = 1;
st->hangover_cnt_he1++;
}
/* Calculate background stationarity */
if ( flag_he1 == 0 && st->first_noise_updt > 0 )
{
if ( snr_sumt > st->bcg_flux )
{
if ( st->bcg_flux_init-- > 0 )
{
if ( snr_sumt > st->bcg_flux+50 )
{
st->bcg_flux = 0.9f * st->bcg_flux + (1-0.9f)*(st->bcg_flux+50);
}
else
{
st->bcg_flux = 0.9f * st->bcg_flux + (1-0.9f)*snr_sumt;
}
}
else
{
if ( snr_sumt > st->bcg_flux+10 )
{
st->bcg_flux = 0.99f * st->bcg_flux + (1-0.99f)*(st->bcg_flux+10);
}
else
{
st->bcg_flux = 0.99f * st->bcg_flux + (1-0.99f)*snr_sumt;
}
}
}
else
{
if ( st->bcg_flux_init-- > 0 )
{
if ( snr_sumt < st->bcg_flux-30 )
{
st->bcg_flux = 0.95f * st->bcg_flux + (1-0.95f)*(st->bcg_flux-30);
}
else
{
st->bcg_flux = 0.95f * st->bcg_flux + (1-0.95f)*snr_sumt;
}
}
else
{
if ( snr_sumt < st->bcg_flux-10 )
{
st->bcg_flux = 0.9992f * st->bcg_flux + (1-0.9992f)*(st->bcg_flux-10);
}
else
{
st->bcg_flux = 0.9992f * st->bcg_flux + (1-0.9992f)*snr_sumt;
}
}
}
if ( st->bcg_flux_init < 0 )
{
st->bcg_flux_init = 0;
}
}
flag = 0;
*localVAD = 0;
if ( snr_sum > thr1 && flag_he1 == 1 ) /* Speech present */
{
flag = 1;
*localVAD = 1;
st->nb_active_frames++; /* Counter of consecutive active speech frames */
if (st->nb_active_frames >= ACTIVE_FRAMES )
{
st->nb_active_frames = ACTIVE_FRAMES;
st->hangover_cnt = 0; /* Reset the counter of hangover frames after at least "active_frames" speech frames */
}
/* inside HO period */
if(st->hangover_cnt < HANGOVER_LONG && st->hangover_cnt != 0 )
{
st->hangover_cnt++;
}
}
else
{
/* Reset the counter of speech frames necessary to start hangover algorithm */
st->nb_active_frames = 0;
if(st->hangover_cnt < HANGOVER_LONG ) /* inside HO period */
{
st->hangover_cnt++;
}
if(st->hangover_cnt <= hangover_short ) /* "hard" hangover */
{
/* send the extra 3 HO frames to NELP */
if ( (lp_snr < th_clean) && (st->Opt_SC_VBR) && (st->hangover_cnt >= 2) )
{
*noisy_speech_HO = 1;
}
if ( (lp_snr >= th_clean) && (st->Opt_SC_VBR) && (st->hangover_cnt >= 2) )
{
*clean_speech_HO = 1;
}
flag = 1;
}
}
/* localVAD and vad_flag for HE-SAD - in parallel with normal localVAD and vad_flag */
*localVAD_HE_SAD = 0;
if( snr_sum_HE_SAD > thr1 && (flag_he1 == 1) ) /* Speech present */
{
*localVAD_HE_SAD = 1;
}
}
/*---------------------------------------------------------------------*
* NB input
* SNR threshold computing
* Hangover control & final VAD decision
*---------------------------------------------------------------------*/
else /* NB input */
{
/* Add localVAD_HE_SAD also for NB operation for use with speech music classifier */
*localVAD_HE_SAD = 0;
if (snr_sum_HE_SAD > thr1 )
{
*localVAD_HE_SAD = 1;
}
*localVAD = 0; /* init needed in NB, otherwise it can be undefined */
if ( snr_sum > thr1 ) /* Speech present */
{
st->nb_active_frames++; /* Counter of consecutive active speech frames */
if (st->nb_active_frames >= ACTIVE_FRAMES )
{
st->nb_active_frames = ACTIVE_FRAMES;
st->hangover_cnt = 0; /* Reset the counter of hangover frames after at least "active_frames" speech frames */
}
*localVAD = 1;
}
else
{
st->nb_active_frames = 0; /* Reset the counter of speech frames necessary to start hangover algorithm */
}
if(st->hangover_cnt < HANGOVER_LONG_NB )
{
st->hangover_cnt++;
if( lp_snr < 19.0f ) /* very low SNR */
{
thr1 -= 5.2f;
}
else if( lp_snr < 35.0f ) /* low SNR */
{
thr1 -= 2.0f;
}
}
if (st->Opt_DTX_ON)
{
if (lp_snr < th_clean)
{
thr2 = thr1 - 1.10f;
}
else
{
thr2 = thr1 - 1.5f;
}
}
else
{
if (lp_snr < th_clean)
{
thr2 = thr1 - 1.3f;
}
else
{
thr2 = thr1 - 1.5f;
}
}
flag = 0;
if ( snr_sum > thr1 ) /* Speech present */
{
flag = 1;
}
if ( (snr_sum < thr1) && (snr_sum > thr2) ) /* Speech present */
{
flag = 1;
*localVAD = 0;
*NB_speech_HO = 1;
}
/* Need to handle the case when switching from WB -> NB */
}
if( st->input_bwidth != NB )
{
*flag_noisy_speech_snr = (lp_snr < TH16_2_NFLAG ); /*original threshold: 35dB*/
}
else
{
*flag_noisy_speech_snr = (lp_snr < TH8_1_NFLAG ); /*original threshold: 20dB, not yet tested!*/
}
/* SC-VBR */
st->vadsnr = snr_sum;
st->vadnoise = thr1;
/* Updates */
st->prim_act_quick = 0.2f * (*localVAD) + (1.0f -0.2f)* st->prim_act_quick;
st->prim_act_slow = 0.01f * (*localVAD) + (1.0f-0.01f)* st->prim_act_slow;
if (st->prim_act_quick <= st->prim_act_slow)
{
st->prim_act = 0.1f * st->prim_act_quick + (1.0f-0.1f)* st->prim_act;
}
else
{
st->prim_act = 0.1f * st->prim_act_slow + (1.0f-0.1f)* st->prim_act;
}
st->prim_act_quick_he = 0.2f * *localVAD_HE_SAD + (1.0f -0.2f)* st->prim_act_quick_he;
st->prim_act_slow_he = 0.01f * *localVAD_HE_SAD + (1.0f-0.01f)* st->prim_act_slow_he;
if (st->prim_act_quick_he <= st->prim_act_slow_he)
{
st->prim_act_he = 0.1f * st->prim_act_quick_he + (1.0f-0.1f)* st->prim_act_he;
}
else
{
st->prim_act_he = 0.1f * st->prim_act_slow_he + (1.0f-0.1f)* st->prim_act_he;
}
if ((st->vad_flag_reg_H & (long) 0x40000L) != 0) /* 0x4000L = 0x01L << 18 */
{
st->vad_flag_cnt_50 = st->vad_flag_cnt_50-1;
}
st->vad_flag_reg_H = (st->vad_flag_reg_H & (long) 0x3fffffffL ) << 1;
if ( ( st->vad_flag_reg_L & (long) 0x40000000L) != 0)
{
st->vad_flag_reg_H = st->vad_flag_reg_H | 0x01L;
}
st->vad_flag_reg_L = (st->vad_flag_reg_L & (long) 0x3fffffffL ) << 1;
if ( flag ) /* should not include the extra DTX hangover */
{
st->vad_flag_reg_L = st->vad_flag_reg_L | 0x01L;
st->vad_flag_cnt_50 = st->vad_flag_cnt_50+1;
}
if ((st->vad_prim_reg & (long) 0x8000L ) != 0) /* 0x8000L = 1L << 15 */
{
st->vad_prim_cnt_16 = st->vad_prim_cnt_16-1;
}
st->vad_prim_reg = (st->vad_prim_reg & (long) 0x3fffffffL ) << 1;
if(*localVAD)
{
st->vad_prim_reg = st->vad_prim_reg | 0x01L;
st->vad_prim_cnt_16 = st->vad_prim_cnt_16+1;
}
return flag;
}