Commit bbe38f4f authored by sekine's avatar sekine
Browse files

Merge branch 'NTT/ROM_and_ITD_202208' into 'main' 

Contribution: Reduction of ROM size and Update of ITD switch for stereo downmix for EVS [Non-BE]

See merge request !98
parents abea63b8 10372244

lib_com/options.h

100755 → 100644
+5 −0
Original line number Diff line number Diff line
@@ -154,6 +154,11 @@ 


#define FIX_I87                                         /*  fix for issue 86: incorrect Ambisonics order set for head rotation in SBA */

#define SBA_ORDER_BITSTREAM                             /* issue 76: Use input sba order for bitstream coding */



/* NTT switches */

#define NTT_UPDATE_ITD_SW                               /* contribution 4: Update of ITD switch in stereo downmix for EVS */

#define NTT_REMOVE_EPS_ROM                              /* contribution 4: Reduction of ROM size in stereo downmix for EVS */



/* ################## End DEVELOPMENT switches ######################### */

/* clang-format on */

#endif

lib_enc/ivas_rom_enc.c

+2 −0
Original line number Diff line number Diff line
@@ -534,6 +534,7 @@ const float ari_bit_estimate_s17_LC[RANGE_N_CONTEXT][RANGE_N_SYMBOLS] = 
 * Stereo downmix to EVS ROM tables

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



#ifndef NTT_REMOVE_EPS_ROM

const float Stereo_dmx_s_wnd_coef_eps_16k[L_FRAME16k * 3 / 4] = {

    0.00000000f, 0.000385506690f, 0.000770864717f, 0.00115592557f, 0.00154054083f, 0.00192456215f, 0.00230784155f, 0.00269023119f, 0.00307158381f, 0.00345175178f,

    0.00383058959f, 0.00420795102f, 0.00458368938f, 0.00495766103f, 0.00532972161f, 0.00569972629f, 0.00606753491f, 0.00643300405f, 0.00679599261f, 0.00715636183f,
@@ -686,6 +687,7 @@ const float Stereo_dmx_s_wnd_coef_eps_48k[L_FRAME48k * 3 / 4] = { 
    -0.00648899190f, -0.00649444433f, -0.00649961829f, -0.00650451379f, -0.00650913082f, -0.00651346892f, -0.00651752809f, -0.00652130833f, -0.00652480870f, -0.00652803015f,

    -0.00653097173f, -0.00653363345f, -0.00653601484f, -0.00653811684f, -0.00653993897f, -0.00654148031f, -0.00654274225f, -0.00654372340f, -0.00654442422f, -0.00654484471f

};

#endif



const float Stereo_dmx_s_wnd_coef_16k[L_FRAME16k >> 4] = {

    0.00154133327f, 0.0138150426f, 0.0380602330f, 0.0736799166f, 0.119797014f, 0.175276011f, 0.238750681f, 0.308658302f, 0.383277327f, 0.460770488f,

lib_enc/ivas_rom_enc.h

+2 −0
Original line number Diff line number Diff line
@@ -120,9 +120,11 @@ extern const uint16_t ECSQ_tab_vals[ECSQ_PARAM_COUNT - 1][1 + ECSQ_TAB_VALS_SIZE 
 * Stereo downmix to EVS ROM tables

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



#ifndef NTT_REMOVE_EPS_ROM

extern const float Stereo_dmx_s_wnd_coef_eps_16k[L_FRAME16k * 3 / 4];

extern const float Stereo_dmx_s_wnd_coef_eps_32k[L_FRAME32k * 3 / 4];

extern const float Stereo_dmx_s_wnd_coef_eps_48k[L_FRAME48k * 3 / 4];

#endif

extern const float Stereo_dmx_s_wnd_coef_16k[L_FRAME16k >> 4];

extern const float Stereo_dmx_s_wnd_coef_32k[L_FRAME32k >> 4];

extern const float Stereo_dmx_s_wnd_coef_48k[L_FRAME48k >> 4];

lib_enc/ivas_stereo_dmx_evs.c

+133 −11
Original line number Diff line number Diff line
@@ -63,6 +63,8 @@ 
#define STEREO_DMX_EVS_DMX_EGY_FORGETTING 0.25f

#define STEREO_DMX_EVS_CORR_FORGETTING    0.78f



#define Q_BAND 0.25f



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

 * Local function prototypes

 *-----------------------------------------------------------------------*/
@@ -150,26 +152,34 @@ static void calc_poc( 
{

    int16_t i, n1, n2;

    int16_t n0, *itdLR;

    const float *c, *s;

#ifndef NTT_REMOVE_EPS_ROM

	const float *c;

#endif

	const float *s;

    float *P;

    float tmp1, tmp2, Lr, Li, Rr, Ri, gamma, igamma, iN;

    float specPOr[L_FRAME48k], specPOi[L_FRAME48k];

    float tmpPOC1[L_FRAME48k], tmpPOC2[L_FRAME48k];

    float rfft_buf[L_FRAME48k];

    int16_t step, bias;

#ifdef NTT_REMOVE_EPS_ROM

    int16_t i_for;

    int16_t cos_step, cos_max;

    float eps_cos, eps_sin, EPS;

#endif



    /* Initialization */

    iN = 1.0f / (float) input_frame;



#ifndef NTT_REMOVE_EPS_ROM

    c = hPOC->sin + ( input_frame >> 2 );

#endif

    s = hPOC->sin;

    P = hPOC->P;

    n0 = input_frame / 2;

    itdLR = hPOC->itdLR;

    igamma = STEREO_DMX_EVS_POC_GAMMA * iN;

    gamma = 1.0f - igamma;

    set_zero( tmpPOC1, L_FRAME48k );

    set_zero( tmpPOC2, L_FRAME48k );



    if ( input_frame == L_FRAME16k )

    {
@@ -185,6 +195,65 @@ static void calc_poc( 
    specPOr[0] = sign( specLr[0] ) * sign( specRr[0] ) * wnd[bias];

    specPOi[0] = 0.0f;



#ifdef NTT_REMOVE_EPS_ROM

    EPS = hPOC->eps;



    if ( input_frame == L_FRAME16k )

    {

        cos_step = 4;

        cos_max = input_frame;

    }

    else /* for 32 kHz & 48 kHz */

    {

        cos_step = 2;

        cos_max = n0;

    }



    for ( i = 1; i < n0 / 2; i++ )

    {

        Lr = specLr[i];

        Li = specLi[i];

        Rr = specRr[i];

        Ri = specRi[i];

        i_for = i * cos_step;

        eps_cos = s[cos_max - i_for] * EPS;

        eps_sin = s[i_for] * EPS;

        Lr += ( specRr[i] * eps_cos + specRi[i] * eps_sin );

        Li += ( -specRr[i] * eps_sin + specRi[i] * eps_cos );

        Rr += ( specLr[i] * eps_cos + specLi[i] * eps_sin );

        Ri += ( -specLr[i] * eps_sin + specLi[i] * eps_cos );

        tmp1 = wnd[i * step + bias] * gamma / ( sqrtf( ( ( Lr * Lr + Li * Li ) ) * ( ( Rr * Rr + Ri * Ri ) ) ) + EPS );



        specPOr[i] = ( Lr * Rr + Li * Ri ) * tmp1;

        specPOi[i] = ( Lr * Ri - Li * Rr ) * tmp1;



        gamma -= igamma;

    }



    for ( i = n0 >> 1; i < n0; i++ )

    {

        Lr = specLr[i];

        Li = specLi[i];

        Rr = specRr[i];

        Ri = specRi[i];



        i_for = ( n0 - i ) * cos_step;

        eps_cos = s[cos_max - i_for] * EPS;

        eps_sin = s[i_for] * EPS;



        Lr += ( -specRr[i] * eps_cos + specRi[i] * eps_sin );

        Li += ( -specRr[i] * eps_sin - specRi[i] * eps_cos );

        Rr += ( -specLr[i] * eps_cos + specLi[i] * eps_sin );

        Ri += ( -specLr[i] * eps_sin - specLi[i] * eps_cos );



        tmp1 = wnd[i * step + bias] * gamma / ( sqrtf( ( ( Lr * Lr + Li * Li ) ) * ( ( Rr * Rr + Ri * Ri ) ) ) + EPS );



        specPOr[i] = ( Lr * Rr + Li * Ri ) * tmp1;

        specPOi[i] = ( Lr * Ri - Li * Rr ) * tmp1;

        gamma -= igamma;

    }

    //end NTT_REMOVE_EPS_ROM

#else

    for ( i = 1; i < n0; i++ )

    {

        Lr = specLr[i];
@@ -204,6 +273,8 @@ static void calc_poc( 


        gamma -= igamma;

    }

#endif //end !NTT_REMOVE_EPS_ROM



    specPOr[n0] = sign( specLr[i] ) * sign( specRr[i] ) * wnd[i * step + bias] * gamma;



    rfft_buf[0] = specPOr[0];
@@ -305,7 +376,7 @@ static float find_poc_peak( 
    itd_cand[0] = itd_cand[1] = 0;

    P = hPOC->P;



    for ( i = 1; i < hPOC->shift_limit; i++ )

    for ( i = 1; i < hPOC->shift_limit; i++ ) /*find peaks of POC P[] with positive and negative ITD */

    {

        if ( P[Lh - i] > Q[0] )

        {
@@ -340,7 +411,7 @@ static float find_poc_peak( 
        Q[n] = ( 1.0f - ( cQ[n] / ( peak_range * 2 + 1 ) + eps2 ) / ( Q[n] + eps2 ) );

        Q[n] = max( Q[n], 0.0f );



        if ( on[n] )

        if ( on[n] ) /*if channel n was active (likely to be preceding) in the previous frame*/

        {

            tmpf = ( 0.3f - 0.2f * (float) abs( itd_cand[n] ) / (float) hPOC->shift_limit ) * peakQ[n];

            if ( Q[n] < tmpf )
@@ -357,7 +428,7 @@ static float find_poc_peak( 


            peakQ[n] = max( peakQ[n], Q[n] );

        }

        else

        else /*if channel n was not active (not likely to be preceding) in the previous frame*/

        {

            tmpf = ( 0.75f - 0.2f * (float) abs( itd_cand[n] ) / (float) hPOC->shift_limit );
@@ -374,6 +445,7 @@ static float find_poc_peak( 
        }

    }



#ifndef NTT_UPDATE_ITD_SW

    if ( on[0] && prev_off[0] )

    {

        *itd = (float) itdLR[0];
@@ -386,8 +458,38 @@ static float find_poc_peak( 
    {

        *itd = ( *itd > 0 ) ? (float) itdLR[0] : (float) itdLR[1];

    }

#else

    if ( ( on[0] && prev_off[0] ) && ( on[1] && prev_off[1] ) ) /*if both channels have newly detected as active (possibility of preceding), select channel by peakness Q[] of POC */

    {

        *itd = ( Q[0] > Q[1] ) ? (float) itdLR[0] : (float) itdLR[1];

    }

    else if ( ( on[0] && prev_off[0] ) && ( Q[0] > ( Q[1] - 0.1 ) ) ) /* if channel 0 becomes active, select channel 0*/

    {

        *itd = (float) itdLR[0];

    }

    else if ( ( on[1] && prev_off[1] ) && ( Q[1] > ( Q[0] - 0.1 ) ) ) /*if channel 1 becomes active, selsect channel 1*/

    {

        *itd = (float) itdLR[1];

    }

    else if ( Q[0] > ( Q[1] + Q_BAND ) ) /* if no status change, use Q[]*/

    {

        *itd = (float) itdLR[0];

    }

    else if ( Q[1] > ( Q[0] + Q_BAND ) ) /* if no status change, use Q[]*/

    {

        *itd = (float) itdLR[1];

    }

    else if ( *itd == 0.0 ) /*if no channels are likely to be preceding, follow the status of the previous frame*/

    {

        *itd = 0;

    }

    else /*follow the status of the previous frame*/

    {

        *itd = ( *itd > 0 ) ? (float) itdLR[0] : (float) itdLR[1];

    }

#endif



    cconfidence = sqrtf( fabsf( Q[0] - Q[1] ) );

    cconfidence = sqrtf( fabsf( Q[0] - Q[1] ) ); /*higher value indicates higher confidence for one preceding channel*/



    return hPOC->confidence = hPOC->confidence * STEREO_DMX_EVS_CORR_FORGETTING + cconfidence * ( 1.0f - STEREO_DMX_EVS_CORR_FORGETTING );

}
@@ -766,6 +868,24 @@ ivas_error stereo_dmx_evs_init_encoder( 
    }

    hStereoDmxEVS->hPOC->eps = 2.0f * EVS_PI / ( (float) input_frame );



#ifdef NTT_REMOVE_EPS_ROM

    if ( input_frame == L_FRAME16k )

    {

        hStereoDmxEVS->hPOC->sin = dft_trigo_32k;

    }

    else if ( input_frame == L_FRAME32k )

    {

        hStereoDmxEVS->hPOC->sin = dft_trigo_32k;

    }

    else if ( input_frame == L_FRAME48k )

    {

        hStereoDmxEVS->hPOC->sin = dft_trigo_48k;

    }

    else

    {

        return IVAS_ERROR( IVAS_ERR_INTERNAL_FATAL, "invalid frame length\n" );

    }

#else

    if ( input_frame == L_FRAME16k )

    {

        hStereoDmxEVS->hPOC->sin = Stereo_dmx_s_wnd_coef_eps_16k;
@@ -782,6 +902,8 @@ ivas_error stereo_dmx_evs_init_encoder( 
    {

        return IVAS_ERROR( IVAS_ERR_INTERNAL_FATAL, "invalid frame length\n" );

    }

#endif



    hStereoDmxEVS->hPOC->confidence = 0.0f;



    *hStereoDmxEVS_out = hStereoDmxEVS;