mirror of
https://github.com/ShaYmez/MMDVM_CM.git
synced 2024-11-10 10:23:28 -05:00
481 lines
13 KiB
C
481 lines
13 KiB
C
|
/*
|
||
|
* Copyright (C) 2010 mbelib Author
|
||
|
* GPG Key ID: 0xEA5EFE2C (9E7A 5527 9CDC EBF7 BF1B D772 4F98 E863 EA5E FE2C)
|
||
|
*
|
||
|
* Permission to use, copy, modify, and/or distribute this software for any
|
||
|
* purpose with or without fee is hereby granted, provided that the above
|
||
|
* copyright notice and this permission notice appear in all copies.
|
||
|
*
|
||
|
* THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES WITH
|
||
|
* REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
|
||
|
* AND FITNESS. IN NO EVENT SHALL ISC BE LIABLE FOR ANY SPECIAL, DIRECT,
|
||
|
* INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
|
||
|
* LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
|
||
|
* OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
|
||
|
* PERFORMANCE OF THIS SOFTWARE.
|
||
|
*/
|
||
|
|
||
|
#include <stdlib.h>
|
||
|
#include <math.h>
|
||
|
#include "ambe.h"
|
||
|
//#include "ambe3600x2250_const.h"
|
||
|
//#include "ambe3600x2400_const.h"
|
||
|
|
||
|
extern const float AmbePlusLtable[];
|
||
|
extern const int AmbePlusVuv[16][8];
|
||
|
extern const int AmbePlusLmprbl[57][4];
|
||
|
extern const float AmbePlusDg[64];
|
||
|
extern const float AmbePlusPRBA24[512][3];
|
||
|
extern const float AmbePlusPRBA58[128][4];
|
||
|
extern const float AmbePlusHOCb5[16][4];
|
||
|
extern const float AmbePlusHOCb6[16][4];
|
||
|
extern const float AmbePlusHOCb7[16][4];
|
||
|
extern const float AmbePlusHOCb8[16][4];
|
||
|
extern const float AmbeW0table[120];
|
||
|
extern const float AmbeLtable[120];
|
||
|
extern const int AmbeVuv[32][8];
|
||
|
extern const int AmbeLmprbl[57][4];
|
||
|
extern const float AmbeDg[32];
|
||
|
extern const float AmbePRBA24[512][3];
|
||
|
extern const float AmbePRBA58[128][4];
|
||
|
extern const float AmbeHOCb5[32][4];
|
||
|
extern const float AmbeHOCb6[16][4];
|
||
|
extern const float AmbeHOCb7[16][4];
|
||
|
extern const float AmbeHOCb8[8][4];
|
||
|
|
||
|
static int
|
||
|
mbe_dequantizeAmbeParms (mbe_parms * cur_mp, mbe_parms * prev_mp, const int *b, int dstar)
|
||
|
{
|
||
|
|
||
|
int ji, i, j, k, l, L, m, am, ak;
|
||
|
int intkl[57];
|
||
|
int b0, b1, b2, b3, b4, b5, b6, b7, b8;
|
||
|
float f0, Cik[5][18], flokl[57], deltal[57];
|
||
|
float Sum42, Sum43, Tl[57], Gm[9], Ri[9], sum, c1, c2;
|
||
|
//char tmpstr[13];
|
||
|
int silence;
|
||
|
int Ji[5], jl;
|
||
|
float deltaGamma, BigGamma;
|
||
|
float unvc, rconst;
|
||
|
|
||
|
b0 = b[0];
|
||
|
b1 = b[1];
|
||
|
b2 = b[2];
|
||
|
b3 = b[3];
|
||
|
b4 = b[4];
|
||
|
b5 = b[5];
|
||
|
b6 = b[6];
|
||
|
b7 = b[7];
|
||
|
b8 = b[8];
|
||
|
|
||
|
silence = 0;
|
||
|
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "\n");
|
||
|
#endif
|
||
|
|
||
|
// copy repeat from prev_mp
|
||
|
cur_mp->repeat = prev_mp->repeat;
|
||
|
|
||
|
if ((b0 >= 120) && (b0 <= 123))
|
||
|
{
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "AMBE Erasure Frame\n");
|
||
|
#endif
|
||
|
return (2);
|
||
|
}
|
||
|
else if ((b0 == 124) || (b0 == 125))
|
||
|
{
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "AMBE Silence Frame\n");
|
||
|
#endif
|
||
|
silence = 1;
|
||
|
cur_mp->w0 = ((float) 2 * M_PI) / (float) 32;
|
||
|
f0 = (float) 1 / (float) 32;
|
||
|
L = 14;
|
||
|
cur_mp->L = 14;
|
||
|
for (l = 1; l <= L; l++)
|
||
|
{
|
||
|
cur_mp->Vl[l] = 0;
|
||
|
}
|
||
|
}
|
||
|
else if ((b0 == 126) || (b0 == 127))
|
||
|
{
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "AMBE Tone Frame\n");
|
||
|
#endif
|
||
|
return (3);
|
||
|
}
|
||
|
|
||
|
if (silence == 0)
|
||
|
{
|
||
|
if (dstar)
|
||
|
f0 = powf(2, (-4.311767578125 - (2.1336e-2 * ((float)b0+0.5))));
|
||
|
else
|
||
|
// w0 from specification document
|
||
|
f0 = AmbeW0table[b0];
|
||
|
cur_mp->w0 = f0 * (float) 2 *M_PI;
|
||
|
// w0 from patent filings
|
||
|
//f0 = powf (2, ((float) b0 + (float) 195.626) / -(float) 45.368);
|
||
|
//cur_mp->w0 = f0 * (float) 2 *M_PI;
|
||
|
}
|
||
|
|
||
|
unvc = (float) 0.2046 / sqrtf (cur_mp->w0);
|
||
|
//unvc = (float) 1;
|
||
|
//unvc = (float) 0.2046 / sqrtf (f0);
|
||
|
|
||
|
// decode L
|
||
|
if (silence == 0)
|
||
|
{
|
||
|
// L from specification document
|
||
|
// lookup L in tabl3
|
||
|
if (dstar)
|
||
|
L = AmbePlusLtable[b0];
|
||
|
else
|
||
|
L = AmbeLtable[b0];
|
||
|
// L formula form patent filings
|
||
|
//L=(int)((float)0.4627 / f0);
|
||
|
cur_mp->L = L;
|
||
|
}
|
||
|
|
||
|
// decode V/UV parameters
|
||
|
for (l = 1; l <= L; l++)
|
||
|
{
|
||
|
// jl from specification document
|
||
|
jl = (int) ((float) l * (float) 16.0 * f0);
|
||
|
// jl from patent filings?
|
||
|
//jl = (int)(((float)l * (float)16.0 * f0) + 0.25);
|
||
|
|
||
|
if (silence == 0)
|
||
|
{
|
||
|
if (dstar)
|
||
|
cur_mp->Vl[l] = AmbePlusVuv[b1][jl];
|
||
|
else
|
||
|
cur_mp->Vl[l] = AmbeVuv[b1][jl];
|
||
|
}
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "jl[%i]:%i Vl[%i]:%i\n", l, jl, l, cur_mp->Vl[l]);
|
||
|
#endif
|
||
|
}
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (atderr, "\nb0:%i w0:%f L:%i b1:%i\n", b0, cur_mp->w0, L, b1);
|
||
|
#endif
|
||
|
if (dstar) {
|
||
|
deltaGamma = AmbePlusDg[b2];
|
||
|
cur_mp->gamma = deltaGamma + ((float) 0.5 * prev_mp->gamma);
|
||
|
} else {
|
||
|
deltaGamma = AmbeDg[b2];
|
||
|
cur_mp->gamma = deltaGamma + ((float) 0.5 * prev_mp->gamma);
|
||
|
}
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "b2: %i, deltaGamma: %f gamma: %f gamma-1: %f\n", b2, deltaGamma, cur_mp->gamma, prev_mp->gamma);
|
||
|
#endif
|
||
|
|
||
|
|
||
|
// decode PRBA vectors
|
||
|
Gm[1] = 0;
|
||
|
|
||
|
if (dstar) {
|
||
|
Gm[2] = AmbePlusPRBA24[b3][0];
|
||
|
Gm[3] = AmbePlusPRBA24[b3][1];
|
||
|
Gm[4] = AmbePlusPRBA24[b3][2];
|
||
|
|
||
|
Gm[5] = AmbePlusPRBA58[b4][0];
|
||
|
Gm[6] = AmbePlusPRBA58[b4][1];
|
||
|
Gm[7] = AmbePlusPRBA58[b4][2];
|
||
|
Gm[8] = AmbePlusPRBA58[b4][3];
|
||
|
} else {
|
||
|
Gm[2] = AmbePRBA24[b3][0];
|
||
|
Gm[3] = AmbePRBA24[b3][1];
|
||
|
Gm[4] = AmbePRBA24[b3][2];
|
||
|
|
||
|
Gm[5] = AmbePRBA58[b4][0];
|
||
|
Gm[6] = AmbePRBA58[b4][1];
|
||
|
Gm[7] = AmbePRBA58[b4][2];
|
||
|
Gm[8] = AmbePRBA58[b4][3];
|
||
|
}
|
||
|
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "b3: %i Gm[2]: %f Gm[3]: %f Gm[4]: %f b4: %i Gm[5]: %f Gm[6]: %f Gm[7]: %f Gm[8]: %f\n", b3, Gm[2], Gm[3], Gm[4], b4, Gm[5], Gm[6], Gm[7], Gm[8]);
|
||
|
#endif
|
||
|
|
||
|
// compute Ri
|
||
|
for (i = 1; i <= 8; i++)
|
||
|
{
|
||
|
sum = 0;
|
||
|
for (m = 1; m <= 8; m++)
|
||
|
{
|
||
|
if (m == 1)
|
||
|
{
|
||
|
am = 1;
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
am = 2;
|
||
|
}
|
||
|
sum = sum + ((float) am * Gm[m] * cosf ((M_PI * (float) (m - 1) * ((float) i - (float) 0.5)) / (float) 8));
|
||
|
}
|
||
|
Ri[i] = sum;
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "R%i: %f ", i, Ri[i]);
|
||
|
#endif
|
||
|
}
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "\n");
|
||
|
#endif
|
||
|
|
||
|
// generate first to elements of each Ci,k block from PRBA vector
|
||
|
rconst = ((float) 1 / ((float) 2 * M_SQRT2));
|
||
|
Cik[1][1] = (float) 0.5 *(Ri[1] + Ri[2]);
|
||
|
Cik[1][2] = rconst * (Ri[1] - Ri[2]);
|
||
|
Cik[2][1] = (float) 0.5 *(Ri[3] + Ri[4]);
|
||
|
Cik[2][2] = rconst * (Ri[3] - Ri[4]);
|
||
|
Cik[3][1] = (float) 0.5 *(Ri[5] + Ri[6]);
|
||
|
Cik[3][2] = rconst * (Ri[5] - Ri[6]);
|
||
|
Cik[4][1] = (float) 0.5 *(Ri[7] + Ri[8]);
|
||
|
Cik[4][2] = rconst * (Ri[7] - Ri[8]);
|
||
|
|
||
|
// decode HOC
|
||
|
|
||
|
// lookup Ji
|
||
|
if (dstar) {
|
||
|
Ji[1] = AmbePlusLmprbl[L][0];
|
||
|
Ji[2] = AmbePlusLmprbl[L][1];
|
||
|
Ji[3] = AmbePlusLmprbl[L][2];
|
||
|
Ji[4] = AmbePlusLmprbl[L][3];
|
||
|
} else {
|
||
|
Ji[1] = AmbeLmprbl[L][0];
|
||
|
Ji[2] = AmbeLmprbl[L][1];
|
||
|
Ji[3] = AmbeLmprbl[L][2];
|
||
|
Ji[4] = AmbeLmprbl[L][3];
|
||
|
}
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "Ji[1]: %i Ji[2]: %i Ji[3]: %i Ji[4]: %i\n", Ji[1], Ji[2], Ji[3], Ji[4]);
|
||
|
fprintf (stderr, "b5: %i b6: %i b7: %i b8: %i\n", b5, b6, b7, b8);
|
||
|
#endif
|
||
|
|
||
|
// Load Ci,k with the values from the HOC tables
|
||
|
// there appear to be a couple typos in eq. 37 so we will just do what makes sense
|
||
|
// (3 <= k <= Ji and k<=6)
|
||
|
for (k = 3; k <= Ji[1]; k++)
|
||
|
{
|
||
|
if (k > 6)
|
||
|
{
|
||
|
Cik[1][k] = 0;
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
if (dstar)
|
||
|
Cik[1][k] = AmbePlusHOCb5[b5][k - 3];
|
||
|
else
|
||
|
Cik[1][k] = AmbeHOCb5[b5][k - 3];
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "C1,%i: %f ", k, Cik[1][k]);
|
||
|
#endif
|
||
|
}
|
||
|
}
|
||
|
for (k = 3; k <= Ji[2]; k++)
|
||
|
{
|
||
|
if (k > 6)
|
||
|
{
|
||
|
Cik[2][k] = 0;
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
if (dstar)
|
||
|
Cik[2][k] = AmbePlusHOCb6[b6][k - 3];
|
||
|
else
|
||
|
Cik[2][k] = AmbeHOCb6[b6][k - 3];
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "C2,%i: %f ", k, Cik[2][k]);
|
||
|
#endif
|
||
|
}
|
||
|
}
|
||
|
for (k = 3; k <= Ji[3]; k++)
|
||
|
{
|
||
|
if (k > 6)
|
||
|
{
|
||
|
Cik[3][k] = 0;
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
if (dstar)
|
||
|
Cik[3][k] = AmbePlusHOCb7[b7][k - 3];
|
||
|
else
|
||
|
Cik[3][k] = AmbeHOCb7[b7][k - 3];
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "C3,%i: %f ", k, Cik[3][k]);
|
||
|
#endif
|
||
|
}
|
||
|
}
|
||
|
for (k = 3; k <= Ji[4]; k++)
|
||
|
{
|
||
|
if (k > 6)
|
||
|
{
|
||
|
Cik[4][k] = 0;
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
if (dstar)
|
||
|
Cik[4][k] = AmbePlusHOCb8[b8][k - 3];
|
||
|
else
|
||
|
Cik[4][k] = AmbeHOCb8[b8][k - 3];
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "C4,%i: %f ", k, Cik[4][k]);
|
||
|
#endif
|
||
|
}
|
||
|
}
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "\n");
|
||
|
#endif
|
||
|
|
||
|
// inverse DCT each Ci,k to give ci,j (Tl)
|
||
|
l = 1;
|
||
|
for (i = 1; i <= 4; i++)
|
||
|
{
|
||
|
ji = Ji[i];
|
||
|
for (j = 1; j <= ji; j++)
|
||
|
{
|
||
|
sum = 0;
|
||
|
for (k = 1; k <= ji; k++)
|
||
|
{
|
||
|
if (k == 1)
|
||
|
{
|
||
|
ak = 1;
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
ak = 2;
|
||
|
}
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "j: %i Cik[%i][%i]: %f ", j, i, k, Cik[i][k]);
|
||
|
#endif
|
||
|
sum = sum + ((float) ak * Cik[i][k] * cosf ((M_PI * (float) (k - 1) * ((float) j - (float) 0.5)) / (float) ji));
|
||
|
}
|
||
|
Tl[l] = sum;
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "Tl[%i]: %f\n", l, Tl[l]);
|
||
|
#endif
|
||
|
l++;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
// determine log2Ml by applying ci,j to previous log2Ml
|
||
|
|
||
|
// fix for when L > L(-1)
|
||
|
if (cur_mp->L > prev_mp->L)
|
||
|
{
|
||
|
for (l = (prev_mp->L) + 1; l <= cur_mp->L; l++)
|
||
|
{
|
||
|
prev_mp->Ml[l] = prev_mp->Ml[prev_mp->L];
|
||
|
prev_mp->log2Ml[l] = prev_mp->log2Ml[prev_mp->L];
|
||
|
}
|
||
|
}
|
||
|
prev_mp->log2Ml[0] = prev_mp->log2Ml[1];
|
||
|
prev_mp->Ml[0] = prev_mp->Ml[1];
|
||
|
|
||
|
// Part 1
|
||
|
Sum43 = 0;
|
||
|
for (l = 1; l <= cur_mp->L; l++)
|
||
|
{
|
||
|
|
||
|
// eq. 40
|
||
|
flokl[l] = ((float) prev_mp->L / (float) cur_mp->L) * (float) l;
|
||
|
intkl[l] = (int) (flokl[l]);
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "flok%i: %f, intk%i: %i ", l, flokl[l], l, intkl[l]);
|
||
|
#endif
|
||
|
// eq. 41
|
||
|
deltal[l] = flokl[l] - (float) intkl[l];
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "delta%i: %f ", l, deltal[l]);
|
||
|
#endif
|
||
|
// eq 43
|
||
|
Sum43 = Sum43 + ((((float) 1 - deltal[l]) * prev_mp->log2Ml[intkl[l]]) + (deltal[l] * prev_mp->log2Ml[intkl[l] + 1]));
|
||
|
}
|
||
|
Sum43 = (((float) 0.65 / (float) cur_mp->L) * Sum43);
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "\n");
|
||
|
fprintf (stderr, "Sum43: %f\n", Sum43);
|
||
|
#endif
|
||
|
|
||
|
// Part 2
|
||
|
Sum42 = 0;
|
||
|
for (l = 1; l <= cur_mp->L; l++)
|
||
|
{
|
||
|
Sum42 += Tl[l];
|
||
|
}
|
||
|
Sum42 = Sum42 / (float) cur_mp->L;
|
||
|
BigGamma = cur_mp->gamma - ((float) 0.5 * (log ((float) cur_mp->L) / log ((float) 2))) - Sum42;
|
||
|
//BigGamma=cur_mp->gamma - ((float)0.5 * log((float)cur_mp->L)) - Sum42;
|
||
|
|
||
|
// Part 3
|
||
|
for (l = 1; l <= cur_mp->L; l++)
|
||
|
{
|
||
|
c1 = ((float) 0.65 * ((float) 1 - deltal[l]) * prev_mp->log2Ml[intkl[l]]);
|
||
|
c2 = ((float) 0.65 * deltal[l] * prev_mp->log2Ml[intkl[l] + 1]);
|
||
|
cur_mp->log2Ml[l] = Tl[l] + c1 + c2 - Sum43 + BigGamma;
|
||
|
// inverse log to generate spectral amplitudes
|
||
|
if (cur_mp->Vl[l] == 1)
|
||
|
{
|
||
|
cur_mp->Ml[l] = exp ((float) 0.693 * cur_mp->log2Ml[l]);
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
cur_mp->Ml[l] = unvc * exp ((float) 0.693 * cur_mp->log2Ml[l]);
|
||
|
}
|
||
|
#ifdef AMBE_DEBUG
|
||
|
fprintf (stderr, "flokl[%i]: %f, intkl[%i]: %i ", l, flokl[l], l, intkl[l]);
|
||
|
fprintf (stderr, "deltal[%i]: %f ", l, deltal[l]);
|
||
|
fprintf (stderr, "prev_mp->log2Ml[%i]: %f\n", l, prev_mp->log2Ml[intkl[l]]);
|
||
|
fprintf (stderr, "BigGamma: %f c1: %f c2: %f Sum43: %f Tl[%i]: %f log2Ml[%i]: %f Ml[%i]: %f\n", BigGamma, c1, c2, Sum43, l, Tl[l], l, cur_mp->log2Ml[l], l, cur_mp->Ml[l]);
|
||
|
#endif
|
||
|
}
|
||
|
|
||
|
return (0);
|
||
|
}
|
||
|
|
||
|
int
|
||
|
mbe_dequantizeAmbeTone(mbe_tone * tone, const int *u)
|
||
|
{
|
||
|
int bitchk1, bitchk2;
|
||
|
int AD, ID1, ID2, ID3, ID4;
|
||
|
bitchk1 = (u[0] >> 6) & 0x3f;
|
||
|
bitchk2 = (u[3] & 0xf);
|
||
|
|
||
|
if ((bitchk1 != 63) || (bitchk2 != 0))
|
||
|
return -1; // Not a valid tone frame
|
||
|
|
||
|
AD = ((u[0] & 0x3f) << 1) + ((u[3] >> 4) & 0x1);
|
||
|
ID1 = ((u[1] & 0xfff) >> 4);
|
||
|
ID2 = ((u[1] & 0xf) << 4) + ((u[2] >> 7) & 0xf);
|
||
|
ID3 = ((u[2] & 0x7f) << 1) + ((u[3] >> 13) & 0x1);
|
||
|
ID4 = ((u[3] & 0x1fe0) >> 5);
|
||
|
|
||
|
if ((ID1 == ID2) && (ID1 == ID3) && (ID1 == ID4) &&
|
||
|
(((ID1 >= 5) && (ID1 <= 122)) || ((ID1 >= 128) && (ID1 <= 163)) || (ID1 == 255))) {
|
||
|
if (tone->ID == ID1) {
|
||
|
tone->AD = AD;
|
||
|
} else {
|
||
|
tone->n = 0;
|
||
|
tone->ID = ID1;
|
||
|
tone->AD = AD;
|
||
|
}
|
||
|
return 0; // valid in-range tone frequency
|
||
|
}
|
||
|
|
||
|
return -1;
|
||
|
}
|
||
|
|
||
|
int
|
||
|
mbe_dequantizeAmbe2400Parms (mbe_parms * cur_mp, mbe_parms * prev_mp, const int *b){
|
||
|
int dstar = 1;
|
||
|
return (mbe_dequantizeAmbeParms (cur_mp, prev_mp, b, dstar));
|
||
|
}
|
||
|
|
||
|
int
|
||
|
mbe_dequantizeAmbe2250Parms (mbe_parms * cur_mp, mbe_parms * prev_mp, const int *b){
|
||
|
int dstar = 0;
|
||
|
return (mbe_dequantizeAmbeParms (cur_mp, prev_mp, b, dstar));
|
||
|
}
|