Commit d2568c77 authored by vaclav's avatar vaclav
Browse files

- Revert "formatting in vad_fx.c"

parent 70490df6

lib_enc/vad_fx.c

+83 −50
Original line number Diff line number Diff line
@@ -13,13 +13,17 @@ 
/*-----------------------------------------------------------------*

 * Local constants

 *-----------------------------------------------------------------*/

/* old fx constants */

#define HANGOVER_LONG    10 /* Hangover for CNG */

#define HANGOVER_LONG_HE 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 */



/* SNR threshold curve constants for WB input */

#define SK16_2_FX 16930 /* Q14  (1.0333f)-> Linear function for clean speech */

#define SC16_2_FX -4608 /* Q8   (-18)*/

#define NK16_2_FX 13529 /* Q15  (.41287)-> Linear function for noisy speech  */

#define NC16_2_FX 3394  /* Q8   (13.259625)*/



/* SNR threshold curve constants for NB input */

#define NK8_1_FX    3509  /*Q15 (0.1071f) Linear function for noisy speech  */

#define NC8_1_FX    4224  /*Q8  (16.5f) */
@@ -28,12 +32,14 @@ 
#define SIGN_THR_FX 40    /*Q4   (2.5f)  Significanse threshold for per band snr calculation */

#define MIN_SNR_FX  2     /*Q4   Minimum snr used for per band snr calculation */





#define THR_M0_FX       3379  /*   Q8 (13.2)  Zero ofset for threshod */

#define THR_K_BG_FX     -8192 /*   Q15(-0.25) Coefficient for dependence on background level */

#define THR_K_BG_OFS_FX 5120  /*   Q8 (20.0f) Zero offset for background level */

#define THR_K_SNR_FX    3277  /*   Q15 (0.1f) Coefficient for dependence on SNR */

#define THR_K_EV_FX     18022 /*   Q15 (0.55) Coefficient for dependence on noise variations */





#define HO_DTX_CLEAN   1  /* Hangover for dtx in clean conditions */

#define HO_DTX_NOISY   10 /* Hangover for dtx in noisy conditions */

#define HO_DTX_NOISY2  4  /* Hangover for dtx in very noisy conditions */
@@ -41,6 +47,7 @@ 


#define ONE_LG10 2466 /*  1.0*log10(2)  in Q13 */





#define HANGOVER_LONG_FX       10 /* Hangover for CNG */

#define HANGOVER_LONG_MUSIC_FX 20 /* Hangover of CNG */

#define HANGOVER_LONG_HE_FX    20 /* Hangover of CNG */
@@ -54,6 +61,7 @@ 
#define TH16_2_NFLAG_FX 8960 /* Q8 (35) */

#define TH8_1_NFLAG_FX  8960 /* Q8 (35) */





#define SNR_OUTLIER_WGHT_1_FX 16384 /* Q14  (1.00)*/

#define SNR_OUTLIER_WGHT_2_FX 16548 /* Q14  (1.01)*/

#define SNR_OUTLIER_WGHT_3_FX 16712 /* Q14  (1.02)*/
@@ -66,12 +74,7 @@ 
#define MAX_SNR_OUTLIER_3_FX   800 /*Q4 (50.0f)*/





/*-----------------------------------------------------------------*

 * vad_snr_log_fx()

 *

 * snr_sum  =   "scale"  * (float)log10( L_snr_sum ) ;

 *-----------------------------------------------------------------*/



/* snr_sum  =   "scale"   * (float)log10( L_snr_sum ) ;*/

/* o: Q8        */

static Word16 vad_snr_log_fx(

    Word32 L_snr, /* i  : Q4        */
@@ -85,7 +88,6 @@ static Word16 vad_snr_log_fx( 
    f_snr = Log2_norm_lc( L_shl( L_snr, e_snr ) );

    e_snr = sub( 30 - 4, e_snr );

    L_tmp = Mpy_32_16( e_snr, f_snr, scale );



    return round_fx( L_shl( L_tmp, 10 ) ); /* Q8 */

}
@@ -167,17 +169,14 @@ void wb_vad_init_fx( 
    move16();

    hVAD->hangover_terminate_flag = 0;

    move16();



    return;

}





/*-----------------------------------------------------------------*

 * sign_thr_snr_acc_fx()

 *

 * accumulate snr_sum with significance thresholds

 *-----------------------------------------------------------------*/



static void sign_thr_snr_acc_ivas_fx(

    Word32 *L_snr_sum, /* o  : q_snr_sum */

    Word16 *q_snr_sum,
@@ -200,11 +199,8 @@ static void sign_thr_snr_acc_ivas_fx( 
    *q_snr_sum = sub( 31, exp_snr_sum );

    move32();

    move32();



    return;

}





static void sign_thr_snr_acc_fx(

    Word32 *L_snr_sum, /* o  : Q4 */

    Word32 L_snr,      /* i  : Q4 */
@@ -225,11 +221,8 @@ static void sign_thr_snr_acc_fx( 
    /* Q4 */

    move32();

    BASOP_SATURATE_WARNING_ON_EVS



    return;

}





/*-----------------------------------------------------------------*

 * dtx_hangover_addition_fx()

 *
@@ -454,7 +447,7 @@ Word16 dtx_hangover_addition_fx( 
 *

 * Voice Activity Detector

 *-----------------------------------------------------------------*/



/* new simplified and harmonized code  */

Word16 wb_vad_fx(

    Encoder_State *st_fx,    /* i/o: encoder state structure                                       */

    const Word32 fr_bands[], /* i  : per band input energy (contains 2 vectors)        Q_new+QSCALE*/
@@ -462,8 +455,8 @@ Word16 wb_vad_fx( 
    Word16 *clean_speech_HO, /* o  : SC-VBR clean speech HO flag                */

    Word16 *NB_speech_HO,    /* o  : SC-VBR NB speech HO flag                   */

    Word16 *snr_sum_he,      /* o  : Output snr_sum as weighted spectral measure*/

    Word16 *localVAD_HE_SAD,       /* o  : HE-SAD flag without hangover               */

    Word16 *flag_noisy_speech_snr, /* o  : encoder detector for noisy speech          */

    Word16 *localVAD_HE_SAD,

    Word16 *flag_noisy_speech_snr, /* o  :   */

    const Word16 Q_new,            /* i  : scaling factor                                            Q0    */

    VAD_HANDLE hVAD,               /* i/o: VAD data handle         */

    NOISE_EST_HANDLE hNoiseEst,    /* i  : Noise estimation handle        */
@@ -507,6 +500,7 @@ Word16 wb_vad_fx( 
    Word16 snr_sum_HE_SAD;   /*Q8 log */

    Word16 sign_thr_HE_SAD, min_snr_HE_SAD;





    Word16 thr1_ol;

    Word32 L_snr_sum_ol;

    Word16 snr_sum_ol; /* Q8 log */
@@ -578,12 +572,11 @@ Word16 wb_vad_fx( 
        last_7k2_coder_type = -1;

        move16();

    }



    /*---------------------------------------------------------------------*

     * set SNR thresholds depending on the input bandwitdh

     *---------------------------------------------------------------------*/



    IF( EQ_16( st_fx->max_band, 19 ) ) /* WB input or SWB input */

    IF( EQ_16( st_fx->max_band, 19 ) )

    /* WB input */ /* or SWB input */

    {

        nk = 3277;

        move16(); /*0.1 Q15 */
@@ -613,8 +606,10 @@ Word16 wb_vad_fx( 


        sign_thr_HE_SAD = 40;

        move16(); /* 2.5f Q4 */

        ;

        min_snr_HE_SAD = 3;

        move16(); /* 0.2f Q4 */

        ;

    }

    ELSE /* NB input */

    {
@@ -635,19 +630,23 @@ Word16 wb_vad_fx( 
        move16(); /* .25 *Q4  MIN_SNR */

        sign_thr_HE_SAD = 42;

        move16(); /* 2.65f Q4 */

        ;

        min_snr_HE_SAD = 1;

        move16(); /* 0.05f Q4 */

        ;

    }



    hangover_short = 0;

    move16();



    /* IF( st_fx->Opt_SC_VBR != 0 ) */

    *noisy_speech_HO = 0;

    move16();

    *clean_speech_HO = 0;

    move16();

    *NB_speech_HO = 0;

    move16();

    /* } */



    /*---------------------------------------------------------------------*

     * compute SNR for each band & total
@@ -668,6 +667,7 @@ Word16 wb_vad_fx( 
        move16();

    }





    IF( snr_idx == 0 )

    {

        stmp = 6;
@@ -681,6 +681,7 @@ Word16 wb_vad_fx( 
        delta4 = 0;

        move16();





        /*vad_thr = 2.4f*lp_snr - 42.2f;

          vad_thr = min(vad_thr,  80 ); */
@@ -748,6 +749,7 @@ Word16 wb_vad_fx( 
    snr_sum_HE_SAD = 0;

    move16();





    FOR( i = st_fx->min_band; i <= st_fx->max_band; i++ )

    {

        ftmp = L_add( *pt1++, 0 );
@@ -833,12 +835,14 @@ Word16 wb_vad_fx( 
        }

        L_snr = L_max( L_snr, 1 * ( 1 << 4 ) ); /* if ( snr[i] < 1 ){snr[i] = 1;}*/





        /* snr[i] = (float)log10(snr[i]);  */

        snr = vad_snr_log_fx( L_snr, ONE_LG10 );



        /* snr_sumt += snr[i];*/

        snr_sumt = add( snr_sumt, shr( snr, 4 ) ); /*Q4 */





        tmp = shl_sat( snr, 5 ); /* Q8 -> Q13 */

        IF( LT_16( i, 2 ) )

        {
@@ -925,6 +929,7 @@ Word16 wb_vad_fx( 
            L_snr = L_shr( snr_tmp, shift_snr ); /*L_snr in Q4*/

        }





        /* conditional snrsum, snr_sum = snr_sum + snr[i];*/

        sign_thr_snr_acc_fx( &L_snr_sum_HE_SAD, L_snr, sign_thr_HE_SAD, min_snr_HE_SAD );

        sign_thr_snr_acc_fx( &L_snr_sum, L_snr, sign_thr, min_snr );
@@ -987,6 +992,7 @@ Word16 wb_vad_fx( 
        }

    }





    /* Separated  SNR_SUM outlier modification  */

    L_snr_sum_ol = L_snr_sum; /* snr_sum_ol = snr_sum;  */
@@ -997,6 +1003,7 @@ Word16 wb_vad_fx( 
    {

        /* Update the total SNR only for WB signals */





        /*    corresponding float section

             if( (accum_ener_L > OUTLIER_THR_1 * accum_ener_H ) || (snr_outlier < MAX_SNR_OUTLIER_1) )

             {
@@ -1016,6 +1023,7 @@ Word16 wb_vad_fx( 
        test();

        IF( LT_32( L_accum_ener_H, Mult_32_16( L_accum_ener_L, INV_OUTLIER_THR_1_FX ) ) /* float::  (accum_ener_L*INV_OUTLIER_THR_1 > accum_ener_H  )  !!!  */

            || LT_32( L_snr_outlier, MAX_SNR_OUTLIER_1_FX ) )



        {

            /* as weight1 is 1.0  we do not need to multiply here ,  i.e. no need to loose any precisison */

            L_snr_sum_ol = L_sub( L_snr_sum_ol, L_snr_outlier ); /*Q4 */
@@ -1054,10 +1062,10 @@ Word16 wb_vad_fx( 
    *snr_sum_he = snr_sum_HE_SAD;

    move16(); /* *snr_sum_he=snr_sum_HE_SAD; */





    /*---------------------------------------------------------------------*

     * compute thr1 for SAD decision

     *---------------------------------------------------------------------*/



    lp_snr = sub( lp_speech_fx, lp_noise_fx ); /*Q8*/



    sub( 0, 0 );
@@ -1095,6 +1103,7 @@ Word16 wb_vad_fx( 
        }

    }





    /*---------------------------------------------------------------------*

     * WB input

     * SNR threshold computing
@@ -1103,6 +1112,7 @@ Word16 wb_vad_fx( 


    IF( NE_16( vad_bwidth_fx, NB ) )

    {



        /* Outlier Detection first calculates thr1_ol and snr_sum_ol instead of

           modyfying thr1 and snr_sum */
@@ -1205,6 +1215,7 @@ Word16 wb_vad_fx( 
            move16();

        }





        IF( GT_16( hVAD->voiced_burst, 3 ) )

        {

            IF( LT_16( hVAD->bcg_flux_fx, 640 ) ) /* Q4 */
@@ -1219,12 +1230,14 @@ Word16 wb_vad_fx( 
            }

        }





        hangover_hd = hangover_hd_tbl[snr_idx];

        move16();



        IF( LT_16( hVAD->bcg_flux_fx, 640 ) )

        {

            hangover_hd = add( shr( hangover_hd, 1 ), 1 );

            // move16();

        }



        /* VAD hangover for he1 */
@@ -1261,6 +1274,7 @@ Word16 wb_vad_fx( 
            move16();

        }





        /* Calculate background stationarity */

        test();

        IF( flag_he1 == 0 && hNoiseEst->first_noise_updt > 0 )
@@ -1405,6 +1419,7 @@ Word16 wb_vad_fx( 
            }

        }





        /* localVAD and vad_flag for HE-SAD - in parallel with normal localVAD and vad_flag */



        *localVAD_HE_SAD = 0;
@@ -1477,6 +1492,7 @@ Word16 wb_vad_fx( 
            }

        }





        thr2 = sub( thr1_nb_mod, 384 ); /*thr2 = thr1 - 1.5f; , clean speech */



        /*    -dtx  condition  dependency   in noisy speech */
@@ -1492,6 +1508,7 @@ Word16 wb_vad_fx( 
            thr2 = sub( thr1_nb_mod, tmp ); /*thr2 = thr1 - [ 1.10 || 1.3 ];*/

        }





        flag = 0;

        move16();

        if ( GT_16( snr_sum, thr1_nb_mod ) ) /* Speech assumed present, even though lowered thr1   */
@@ -1500,6 +1517,7 @@ Word16 wb_vad_fx( 
            move16();

        }





        test();

        IF( ( LT_16( snr_sum, thr1_nb_mod ) ) && ( GT_16( snr_sum, thr2 ) ) ) /* Speech present */

        {
@@ -1518,6 +1536,7 @@ Word16 wb_vad_fx( 
        move16(); /* needed for  st_fx->vadnoise_fx update below */

    } /* end of NB */





    /* *flag_noisy_speech_snr is a  Word8 parameter  */

    *flag_noisy_speech_snr = 0; /*_DIFF_FLOAT_FIX_IVAS -> this initialisation is not done here in float */

    move16();
@@ -1628,6 +1647,7 @@ Word16 wb_vad_fx( 
        move16();

    }





    IF( L_and( hVAD->vad_prim_reg, (Word32) 0x8000L ) != 0 ) /* 0x8000L = 1L << 15 */

    {

        hVAD->vad_prim_cnt_16 = sub( hVAD->vad_prim_cnt_16, 1 );
@@ -1655,8 +1675,8 @@ Word16 wb_vad_ivas_fx( 
    Word16 *clean_speech_HO, /* o  : SC-VBR clean speech HO flag                */

    Word16 *NB_speech_HO,    /* o  : SC-VBR NB speech HO flag                   */

    Word16 *snr_sum_he,      /* o  : Output snr_sum as weighted spectral measure*/

    Word16 *localVAD_HE_SAD,       /* o  : HE-SAD flag without hangover               */

    Word16 *flag_noisy_speech_snr, /* o  : encoder detector for noisy speech          */

    Word16 *localVAD_HE_SAD,

    Word16 *flag_noisy_speech_snr, /* o  :   */

    VAD_HANDLE hVAD,               /* i/o: VAD data handle         */

    NOISE_EST_HANDLE hNoiseEst,    /* i  : Noise estimation handle        */

    Word16 lp_speech_fx,           /* i  : long term active speech energy average     Q8   */
@@ -1787,12 +1807,11 @@ Word16 wb_vad_ivas_fx( 
        last_7k2_coder_type = -1;

        move16();

    }



    /*---------------------------------------------------------------------*

     * set SNR thresholds depending on the input bandwitdh

     *---------------------------------------------------------------------*/



    IF( EQ_16( st_fx->max_band, 19 ) ) /* WB input or SWB input */

    IF( EQ_16( st_fx->max_band, 19 ) )

    /* WB input */ /* or SWB input */

    {

        nk = 3277;

        move16(); /*0.1 Q15 */
@@ -1827,6 +1846,7 @@ Word16 wb_vad_ivas_fx( 
    }

    ELSE /* NB input */

    {

        // move16();

        nk = 3277;

        move16(); /*  0.1 Q15 */

        nc = 268435456;
@@ -1879,6 +1899,7 @@ Word16 wb_vad_ivas_fx( 
        move16();

    }





    IF( snr_idx == 0 )

    {

        stmp = 6;
@@ -1892,6 +1913,7 @@ Word16 wb_vad_ivas_fx( 
        delta4 = 0;

        move32(); /*0.0f in Q16 */





        /*vad_thr = 2.4f*lp_snr - 42.2f;

          vad_thr = min(vad_thr,  80 ); */
@@ -2249,6 +2271,7 @@ Word16 wb_vad_ivas_fx( 
        test();

        IF( LT_32( L_accum_ener_H, Mult_32_16( L_accum_ener_L, INV_OUTLIER_THR_1_FX ) ) /* float::  (accum_ener_L*INV_OUTLIER_THR_1 > accum_ener_H  )  !!!  */

            || LT_32( L_snr_outlier_Q4, MAX_SNR_OUTLIER_1_FX ) )



        {

            /* snr_sum_ol = SNR_OUTLIER_WGHT_1 * ( snr_sum_ol - snr_outlier );

              As SNR_OUTLIER_WGHT_1 is 1.0f  we do not need to multiply here ,  i.e. no need to loose any precisison */
@@ -2295,10 +2318,10 @@ Word16 wb_vad_ivas_fx( 
    *snr_sum_he = snr_sum_HE_SAD;

    move16(); /* *snr_sum_he=snr_sum_HE_SAD; */





    /*---------------------------------------------------------------------*

     * compute thr1 for SAD decision

     *---------------------------------------------------------------------*/



    lp_snr = sub( lp_speech_fx, lp_noise_fx ); /*Q8*/



    IF( LT_16( lp_snr, hNoiseEst->sign_dyn_lp_fx ) )
@@ -2346,6 +2369,7 @@ Word16 wb_vad_ivas_fx( 
        }

    }





    /*---------------------------------------------------------------------*

     * WB input

     * SNR threshold computing
@@ -2354,6 +2378,7 @@ Word16 wb_vad_ivas_fx( 


    IF( vad_bwidth_fx != NB )

    {



        /* Outlier Detection first calculates thr1_ol and snr_sum_ol instead of

           modyfying thr1 and snr_sum */
@@ -2465,6 +2490,7 @@ Word16 wb_vad_ivas_fx( 
            move16();

        }





        IF( GT_16( hVAD->voiced_burst, 3 ) )

        {

            IF( LT_16( hVAD->bcg_flux_fx, 640 ) ) /* Q4 */
@@ -2479,6 +2505,7 @@ Word16 wb_vad_ivas_fx( 
            }

        }





        hangover_hd = hangover_hd_tbl[snr_idx];

        move16();
@@ -2521,6 +2548,7 @@ Word16 wb_vad_ivas_fx( 
            move16();

        }





        /* Calculate background stationarity */

        test();

        IF( flag_he1 == 0 && hNoiseEst->first_noise_updt > 0 )
@@ -2739,6 +2767,7 @@ Word16 wb_vad_ivas_fx( 
            }

        }





        thr2 = sub( thr1_nb_mod, 384 ); /*thr2 = thr1 - 1.5f; , clean speech */



        /*    -dtx  condition  dependency   in noisy speech */
@@ -2754,6 +2783,7 @@ Word16 wb_vad_ivas_fx( 
            thr2 = sub( thr1_nb_mod, tmp ); /*thr2 = thr1 - [ 1.10 || 1.3 ];*/

        }





        flag = 0;

        move16();

        IF( GT_16( snr_sum, thr1_nb_mod ) ) /* Speech assumed present, even though lowered thr1   */
@@ -2762,6 +2792,7 @@ Word16 wb_vad_ivas_fx( 
            move16();

        }





        test();

        IF( ( LT_16( snr_sum, thr1_nb_mod ) ) && ( GT_16( snr_sum, thr2 ) ) ) /* Speech present */

        {
@@ -2780,6 +2811,7 @@ Word16 wb_vad_ivas_fx( 
        move16(); /* needed for  st_fx->vadnoise_fx update below */

    } /* end of NB */





    /* *flag_noisy_speech_snr is a  Word8 parameter  */

    *flag_noisy_speech_snr = 0; /*_DIFF_FLOAT_FIX_IVAS -> this initialisation is not done here in float */

    move16();
@@ -2891,6 +2923,7 @@ Word16 wb_vad_ivas_fx( 
        move16();

    }





    IF( L_and( hVAD->vad_prim_reg, (Word32) 0x8000L ) != 0 ) /* 0x8000L = 1L << 15 */

    {

        hVAD->vad_prim_cnt_16 = sub( hVAD->vad_prim_cnt_16, 1 ); /*Q0*/