You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

502 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 <stdio.h>
#include <math.h>
#include "mbelib.h"
#include "mbelib_const.h"
/**
* \return A pseudo-random float between [0.0, 1.0].
* See http://www.azillionmonkeys.com/qed/random.html for further improvements
*/
static float
mbe_rand()
{
return ((float) rand () / (float) RAND_MAX);
}
/**
* \return A pseudo-random float between [-pi, +pi].
*/
static float
mbe_rand_phase()
{
return mbe_rand() * (((float)M_PI) * 2.0F) - ((float)M_PI);
}
void
mbe_printVersion (char *str)
{
sprintf (str, "%s", MBELIB_VERSION);
}
void
mbe_moveMbeParms (mbe_parms * cur_mp, mbe_parms * prev_mp)
{
int l;
prev_mp->w0 = cur_mp->w0;
prev_mp->L = cur_mp->L;
prev_mp->K = cur_mp->K; // necessary?
prev_mp->Ml[0] = (float) 0;
prev_mp->gamma = cur_mp->gamma;
prev_mp->repeat = cur_mp->repeat;
for (l = 0; l <= 56; l++)
{
prev_mp->Ml[l] = cur_mp->Ml[l];
prev_mp->Vl[l] = cur_mp->Vl[l];
prev_mp->log2Ml[l] = cur_mp->log2Ml[l];
prev_mp->PHIl[l] = cur_mp->PHIl[l];
prev_mp->PSIl[l] = cur_mp->PSIl[l];
}
}
void
mbe_useLastMbeParms (mbe_parms * cur_mp, mbe_parms * prev_mp)
{
int l;
cur_mp->w0 = prev_mp->w0;
cur_mp->L = prev_mp->L;
cur_mp->K = prev_mp->K; // necessary?
cur_mp->Ml[0] = (float) 0;
cur_mp->gamma = prev_mp->gamma;
cur_mp->repeat = prev_mp->repeat;
for (l = 0; l <= 56; l++)
{
cur_mp->Ml[l] = prev_mp->Ml[l];
cur_mp->Vl[l] = prev_mp->Vl[l];
cur_mp->log2Ml[l] = prev_mp->log2Ml[l];
cur_mp->PHIl[l] = prev_mp->PHIl[l];
cur_mp->PSIl[l] = prev_mp->PSIl[l];
}
}
void
mbe_initMbeParms (mbe_parms * cur_mp, mbe_parms * prev_mp, mbe_parms * prev_mp_enhanced)
{
int l;
prev_mp->w0 = 0.09378;
prev_mp->L = 30;
prev_mp->K = 10;
prev_mp->gamma = (float) 0;
for (l = 0; l <= 56; l++)
{
prev_mp->Ml[l] = (float) 0;
prev_mp->Vl[l] = 0;
prev_mp->log2Ml[l] = (float) 0; // log2 of 1 == 0
prev_mp->PHIl[l] = (float) 0;
prev_mp->PSIl[l] = (M_PI / (float) 2);
}
prev_mp->repeat = 0;
mbe_moveMbeParms (prev_mp, cur_mp);
mbe_moveMbeParms (prev_mp, prev_mp_enhanced);
}
void
mbe_spectralAmpEnhance (mbe_parms * cur_mp)
{
float Rm0, Rm1, R2m0, R2m1, Wl[57];
int l;
float sum, gamma, M;
Rm0 = 0;
Rm1 = 0;
for (l = 1; l <= cur_mp->L; l++)
{
Rm0 = Rm0 + (cur_mp->Ml[l] * cur_mp->Ml[l]);
Rm1 = Rm1 + ((cur_mp->Ml[l] * cur_mp->Ml[l]) * cosf (cur_mp->w0 * (float) l));
}
R2m0 = (Rm0*Rm0);
R2m1 = (Rm1*Rm1);
for (l = 1; l <= cur_mp->L; l++)
{
if (cur_mp->Ml[l] != 0)
{
Wl[l] = sqrtf (cur_mp->Ml[l]) * powf ((((float) 0.96 * M_PI * ((R2m0 + R2m1) - ((float) 2 * Rm0 * Rm1 * cosf (cur_mp->w0 * (float) l)))) / (cur_mp->w0 * Rm0 * (R2m0 - R2m1))), (float) 0.25);
if ((8 * l) <= cur_mp->L)
{
}
else if (Wl[l] > 1.2)
{
cur_mp->Ml[l] = 1.2 * cur_mp->Ml[l];
}
else if (Wl[l] < 0.5)
{
cur_mp->Ml[l] = 0.5 * cur_mp->Ml[l];
}
else
{
cur_mp->Ml[l] = Wl[l] * cur_mp->Ml[l];
}
}
}
// generate scaling factor
sum = 0;
for (l = 1; l <= cur_mp->L; l++)
{
M = cur_mp->Ml[l];
if (M < 0)
{
M = -M;
}
sum += (M*M);
}
if (sum == 0)
{
gamma = (float) 1.0;
}
else
{
gamma = sqrtf (Rm0 / sum);
}
// apply scaling factor
for (l = 1; l <= cur_mp->L; l++)
{
cur_mp->Ml[l] = gamma * cur_mp->Ml[l];
}
}
void
mbe_synthesizeSilencef (float *aout_buf)
{
int n;
float *aout_buf_p;
aout_buf_p = aout_buf;
for (n = 0; n < 160; n++)
{
*aout_buf_p = (float) 0;
aout_buf_p++;
}
}
void
mbe_synthesizeSilence (short *aout_buf)
{
int n;
short *aout_buf_p;
aout_buf_p = aout_buf;
for (n = 0; n < 160; n++)
{
*aout_buf_p = (short) 0;
aout_buf_p++;
}
}
void
mbe_synthesizeSpeechf (float *aout_buf, mbe_parms * cur_mp, mbe_parms * prev_mp, int uvquality)
{
int i, l, n, maxl;
float *Ss, loguvquality;
float C1, C2, C3, C4;
//float deltaphil, deltawl, thetaln, aln;
int numUv;
float cw0, pw0, cw0l, pw0l;
float uvsine, uvrand, uvthreshold, uvthresholdf;
float uvstep, uvoffset;
float qfactor;
float rphase[64], rphase2[64];
const int N = 160;
uvthresholdf = (float) 2700;
uvthreshold = ((uvthresholdf * M_PI) / (float) 4000);
// voiced/unvoiced/gain settings
uvsine = (float) 1.3591409 *M_E;
uvrand = (float) 2.0;
if ((uvquality < 1) || (uvquality > 64))
{
printf ("\nmbelib: Error - uvquality must be within the range 1 - 64, setting to default value of 3\n");
uvquality = 3;
}
// calculate loguvquality
if (uvquality == 1)
{
loguvquality = (float) 1 / M_E;
}
else
{
loguvquality = log ((float) uvquality) / (float) uvquality;
}
// calculate unvoiced step and offset values
uvstep = (float) 1.0 / (float) uvquality;
qfactor = loguvquality;
uvoffset = (uvstep * (float) (uvquality - 1)) / (float) 2;
// count number of unvoiced bands
numUv = 0;
for (l = 1; l <= cur_mp->L; l++)
{
if (cur_mp->Vl[l] == 0)
{
numUv++;
}
}
cw0 = cur_mp->w0;
pw0 = prev_mp->w0;
// init aout_buf
Ss = aout_buf;
for (n = 0; n < N; n++)
{
*Ss = (float) 0;
Ss++;
}
// eq 128 and 129
if (cur_mp->L > prev_mp->L)
{
maxl = cur_mp->L;
for (l = prev_mp->L + 1; l <= maxl; l++)
{
prev_mp->Ml[l] = (float) 0;
prev_mp->Vl[l] = 1;
}
}
else
{
maxl = prev_mp->L;
for (l = cur_mp->L + 1; l <= maxl; l++)
{
cur_mp->Ml[l] = (float) 0;
cur_mp->Vl[l] = 1;
}
}
// update phil from eq 139,140
for (l = 1; l <= 56; l++)
{
cur_mp->PSIl[l] = prev_mp->PSIl[l] + ((pw0 + cw0) * ((float) (l * N) / (float) 2));
if (l <= (int) (cur_mp->L / 4))
{
cur_mp->PHIl[l] = cur_mp->PSIl[l];
}
else
{
cur_mp->PHIl[l] = cur_mp->PSIl[l] + ((numUv * mbe_rand_phase()) / cur_mp->L);
}
}
for (l = 1; l <= maxl; l++)
{
cw0l = (cw0 * (float) l);
pw0l = (pw0 * (float) l);
if ((cur_mp->Vl[l] == 0) && (prev_mp->Vl[l] == 1))
{
Ss = aout_buf;
// init random phase
for (i = 0; i < uvquality; i++)
{
rphase[i] = mbe_rand_phase();
}
for (n = 0; n < N; n++)
{
C1 = 0;
// eq 131
C1 = Ws[n + N] * prev_mp->Ml[l] * cosf ((pw0l * (float) n) + prev_mp->PHIl[l]);
C3 = 0;
// unvoiced multisine mix
for (i = 0; i < uvquality; i++)
{
C3 = C3 + cosf ((cw0 * (float) n * ((float) l + ((float) i * uvstep) - uvoffset)) + rphase[i]);
if (cw0l > uvthreshold)
{
C3 = C3 + ((cw0l - uvthreshold) * uvrand * mbe_rand());
}
}
C3 = C3 * uvsine * Ws[n] * cur_mp->Ml[l] * qfactor;
*Ss = *Ss + C1 + C3;
Ss++;
}
}
else if ((cur_mp->Vl[l] == 1) && (prev_mp->Vl[l] == 0))
{
Ss = aout_buf;
// init random phase
for (i = 0; i < uvquality; i++)
{
rphase[i] = mbe_rand_phase();
}
for (n = 0; n < N; n++)
{
C1 = 0;
// eq 132
C1 = Ws[n] * cur_mp->Ml[l] * cosf ((cw0l * (float) (n - N)) + cur_mp->PHIl[l]);
C3 = 0;
// unvoiced multisine mix
for (i = 0; i < uvquality; i++)
{
C3 = C3 + cosf ((pw0 * (float) n * ((float) l + ((float) i * uvstep) - uvoffset)) + rphase[i]);
if (pw0l > uvthreshold)
{
C3 = C3 + ((pw0l - uvthreshold) * uvrand * mbe_rand());
}
}
C3 = C3 * uvsine * Ws[n + N] * prev_mp->Ml[l] * qfactor;
*Ss = *Ss + C1 + C3;
Ss++;
}
}
// else if (((cur_mp->Vl[l] == 1) || (prev_mp->Vl[l] == 1)) && ((l >= 8) || (fabsf (cw0 - pw0) >= ((float) 0.1 * cw0))))
else if ((cur_mp->Vl[l] == 1) || (prev_mp->Vl[l] == 1))
{
Ss = aout_buf;
for (n = 0; n < N; n++)
{
C1 = 0;
// eq 133-1
C1 = Ws[n + N] * prev_mp->Ml[l] * cosf ((pw0l * (float) n) + prev_mp->PHIl[l]);
C2 = 0;
// eq 133-2
C2 = Ws[n] * cur_mp->Ml[l] * cosf ((cw0l * (float) (n - N)) + cur_mp->PHIl[l]);
*Ss = *Ss + C1 + C2;
Ss++;
}
}
/*
// expensive and unnecessary?
else if ((cur_mp->Vl[l] == 1) || (prev_mp->Vl[l] == 1))
{
Ss = aout_buf;
// eq 137
deltaphil = cur_mp->PHIl[l] - prev_mp->PHIl[l] - (((pw0 + cw0) * (float) (l * N)) / (float) 2);
// eq 138
deltawl = ((float) 1 / (float) N) * (deltaphil - ((float) 2 * M_PI * (int) ((deltaphil + M_PI) / (M_PI * (float) 2))));
for (n = 0; n < N; n++)
{
// eq 136
thetaln = prev_mp->PHIl[l] + ((pw0l + deltawl) * (float) n) + (((cw0 - pw0) * ((float) (l * n * n)) / (float) (2 * N)));
// eq 135
aln = prev_mp->Ml[l] + (((float) n / (float) N) * (cur_mp->Ml[l] - prev_mp->Ml[l]));
// eq 134
*Ss = *Ss + (aln * cosf (thetaln));
Ss++;
}
}
*/
else
{
Ss = aout_buf;
// init random phase
for (i = 0; i < uvquality; i++)
{
rphase[i] = mbe_rand_phase();
}
// init random phase
for (i = 0; i < uvquality; i++)
{
rphase2[i] = mbe_rand_phase();
}
for (n = 0; n < N; n++)
{
C3 = 0;
// unvoiced multisine mix
for (i = 0; i < uvquality; i++)
{
C3 = C3 + cosf ((pw0 * (float) n * ((float) l + ((float) i * uvstep) - uvoffset)) + rphase[i]);
if (pw0l > uvthreshold)
{
C3 = C3 + ((pw0l - uvthreshold) * uvrand * mbe_rand());
}
}
C3 = C3 * uvsine * Ws[n + N] * prev_mp->Ml[l] * qfactor;
C4 = 0;
// unvoiced multisine mix
for (i = 0; i < uvquality; i++)
{
C4 = C4 + cosf ((cw0 * (float) n * ((float) l + ((float) i * uvstep) - uvoffset)) + rphase2[i]);
if (cw0l > uvthreshold)
{
C4 = C4 + ((cw0l - uvthreshold) * uvrand * mbe_rand());
}
}
C4 = C4 * uvsine * Ws[n] * cur_mp->Ml[l] * qfactor;
*Ss = *Ss + C3 + C4;
Ss++;
}
}
}
}
void
mbe_synthesizeSpeech (short *aout_buf, mbe_parms * cur_mp, mbe_parms * prev_mp, int uvquality)
{
float float_buf[160];
mbe_synthesizeSpeechf (float_buf, cur_mp, prev_mp, uvquality);
mbe_floattoshort (float_buf, aout_buf);
}
void
mbe_floattoshort (float *float_buf, short *aout_buf)
{
short *aout_buf_p;
float *float_buf_p;
int i, again;
float audio;
again = 7;
aout_buf_p = aout_buf;
float_buf_p = float_buf;
for (i = 0; i < 160; i++)
{
audio = again * *float_buf_p;
if (audio > 32760)
{
#ifdef MBE_DEBUG
printf ("audio clip: %f\n", audio);
#endif
audio = 32760;
}
else if (audio < -32760)
{
#ifdef MBE_DEBUG
printf ("audio clip: %f\n", audio);
#endif
audio = -32760;
}
*aout_buf_p = (short) (audio);
aout_buf_p++;
float_buf_p++;
}
}