ref: 8b6936294846c08a320e4dc9a95d2810068e1a33
dir: /libfaac/ltp.c/
/*
* FAAC - Freeware Advanced Audio Coder
* Copyright (C) 2001 Menno Bakker
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* $Id: ltp.c,v 1.6 2001/06/08 18:01:09 menno Exp $
*/
#include <stdio.h>
#include <math.h>
#include "frame.h"
#include "coder.h"
#include "ltp.h"
#include "tns.h"
#include "filtbank.h"
#include "util.h"
/* short double_to_int(double sig_in); */
#define double_to_int(sig_in) \
((sig_in) > 32767 ? 32767 : ( \
(sig_in) < -32768 ? -32768 : (sig_in)))
/* Purpose: Codebook for LTP weight coefficients. */
static double codebook[CODESIZE] =
{
0.570829,
0.696616,
0.813004,
0.911304,
0.984900,
1.067894,
1.194601,
1.369533
};
static double snr_pred(double *mdct_in, double *mdct_pred, int *sfb_flag, int *sfb_offset,
int block_type, int side_info, int num_of_sfb)
{
int i, j, flen;
double snr_limit;
double num_bit, snr[NSFB_LONG];
double temp1, temp2;
double energy[BLOCK_LEN_LONG], snr_p[BLOCK_LEN_LONG];
if (block_type != ONLY_SHORT_WINDOW)
{
flen = BLOCK_LEN_LONG;
snr_limit = 1.e-30;
} else {
flen = BLOCK_LEN_SHORT;
snr_limit = 1.e-20;
}
for (i = 0; i < flen; i++)
{
energy[i] = mdct_in[i] * mdct_in[i];
snr_p[i] = (mdct_in[i] - mdct_pred[i]) * (mdct_in[i] - mdct_pred[i]);
}
num_bit = 0.0;
for (i = 0; i < num_of_sfb; i++)
{
temp1 = 0.0;
temp2 = 0.0;
for (j = sfb_offset[i]; j < sfb_offset[i + 1]; j++)
{
temp1 += energy[j];
temp2 += snr_p[j];
}
if (temp2 < snr_limit)
temp2 = snr_limit;
if (temp1 > 1.e-20)
snr[i] = -10. * log10 (temp2 / temp1);
else
snr[i] = 0.0;
sfb_flag[i] = 1;
if (block_type != ONLY_SHORT_WINDOW)
{
if (snr[i] <= 0.0)
{
sfb_flag[i] = 0;
for (j = sfb_offset[i]; j < sfb_offset[i + 1]; j++)
mdct_pred[j] = 0.0;
} else {
num_bit += snr[i] / 6. * (sfb_offset[i + 1] - sfb_offset[i]);
}
}
}
if (num_bit < side_info)
{
num_bit = 0.0;
for (j = 0; j < flen; j++)
mdct_pred[j] = 0.0;
for (i = 0; i < num_of_sfb; i++)
sfb_flag[i] = 0;
} else {
num_bit -= side_info;
}
return (num_bit);
}
static void prediction(double *buffer, double *predicted_samples, double *weight, int lag,
int flen)
{
int i, offset;
int num_samples;
offset = NOK_LT_BLEN - flen / 2 - lag;
num_samples = flen;
if(NOK_LT_BLEN - offset < flen)
num_samples = NOK_LT_BLEN - offset;
for(i = 0; i < num_samples; i++)
predicted_samples[i] = *weight * buffer[offset++];
for( ; i < flen; i++)
predicted_samples[i] = 0.0;
}
static void w_quantize(double *freq, int *ltp_idx)
{
int i;
double dist, low;
low = 1.0e+10;
dist = 0.0;
for (i = 0; i < CODESIZE; i++)
{
dist = (*freq - codebook[i]) * (*freq - codebook[i]);
if (dist < low)
{
low = dist;
*ltp_idx = i;
}
}
*freq = codebook[*ltp_idx];
}
static int pitch(double *sb_samples, double *x_buffer, int flen, int lag0, int lag1,
double *predicted_samples, double *gain, int *cb_idx)
{
int i, j, delay;
double corr1, corr2, lag_corr, corrtmp;
double p_max, energy, lag_energy;
/*
* Below is a figure illustrating how the lag and the
* samples in the buffer relate to each other.
*
* ------------------------------------------------------------------
* | | | | |
* | slot 1 | 2 | 3 | 4 |
* | | | | |
* ------------------------------------------------------------------
*
* lag = 0 refers to the end of slot 4 and lag = DELAY refers to the end
* of slot 2. The start of the predicted frame is then obtained by
* adding the length of the frame to the lag. Remember that slot 4 doesn't
* actually exist, since it is always filled with zeros.
*
* The above short explanation was for long blocks. For short blocks the
* zero lag doesn't refer to the end of slot 4 but to the start of slot
* 4 - the frame length of a short block.
*
* Some extra code is then needed to handle those lag values that refer
* to slot 4.
*/
p_max = 0.0;
lag_corr = lag_energy = 0.0;
delay = lag0;
energy = 0.0;
corr1 = 0.0;
for (j = lag0; j < lag1; j++)
{
corr1 += x_buffer[NOK_LT_BLEN - j - 1] * sb_samples[flen - j - 1];
energy += x_buffer[NOK_LT_BLEN - j - 1] * x_buffer[NOK_LT_BLEN - j - 1];
}
corrtmp=corr1;
if (energy != 0.0)
corr2 = corr1 / sqrt(energy);
else
corr2 = 0.0;
if (p_max < corr2)
{
p_max = corr2;
delay = 0;
lag_corr = corr1;
lag_energy = energy;
}
/* Find the lag. */
for (i = lag0 + 1; i < lag1; i++)
{
energy -= x_buffer[NOK_LT_BLEN - i] * x_buffer[NOK_LT_BLEN - i];
energy += x_buffer[NOK_LT_BLEN - i - flen] * x_buffer[NOK_LT_BLEN - i - flen];
corr1 = corrtmp;
corr1 -= x_buffer[NOK_LT_BLEN - i] * sb_samples[flen - 1];
corr1 += x_buffer[NOK_LT_BLEN - i - flen] * sb_samples[0];
corrtmp = corr1;
if (energy != 0.0)
corr2 = corr1 / sqrt(energy);
else
corr2 = 0.0;
if (p_max < corr2)
{
p_max = corr2;
delay = i;
lag_corr = corr1;
lag_energy = energy;
}
}
/* Compute the gain. */
if(lag_energy != 0.0)
*gain = lag_corr / (1.010 * lag_energy);
else
*gain = 0.0;
/* Quantize the gain. */
w_quantize(gain, cb_idx);
/* Get the predicted signal. */
prediction(x_buffer, predicted_samples, gain, delay, flen);
return (delay);
}
static double ltp_enc_tf(faacEncHandle hEncoder,
CoderInfo *coderInfo, double *p_spectrum, double *predicted_samples,
double *mdct_predicted, int *sfb_offset,
int num_of_sfb, int last_band, int side_info,
int *sfb_prediction_used, TnsInfo *tnsInfo)
{
double bit_gain;
/* Transform prediction to frequency domain. */
FilterBank(hEncoder, coderInfo, predicted_samples, mdct_predicted,
NULL, MNON_OVERLAPPED);
/* Apply TNS analysis filter to the predicted spectrum. */
if(tnsInfo != NULL)
TnsEncodeFilterOnly(tnsInfo, num_of_sfb, num_of_sfb, coderInfo->block_type, sfb_offset,
mdct_predicted);
/* Get the prediction gain. */
bit_gain = snr_pred(p_spectrum, mdct_predicted, sfb_prediction_used,
sfb_offset, side_info, last_band, coderInfo->nr_of_sfb);
return (bit_gain);
}
void LtpInit(faacEncHandle hEncoder)
{
int i;
unsigned int channel;
for (channel = 0; channel < hEncoder->numChannels; channel++) {
LtpInfo *ltpInfo = &(hEncoder->coderInfo[channel].ltpInfo);
ltpInfo->buffer = AllocMemory(NOK_LT_BLEN * sizeof(double));
ltpInfo->mdct_predicted = AllocMemory(2*BLOCK_LEN_LONG*sizeof(double));
ltpInfo->time_buffer = AllocMemory(BLOCK_LEN_LONG*sizeof(double));
ltpInfo->ltp_overlap_buffer = AllocMemory(BLOCK_LEN_LONG*sizeof(double));
for (i = 0; i < NOK_LT_BLEN; i++)
ltpInfo->buffer[i] = 0;
ltpInfo->weight_idx = 0;
for(i = 0; i < MAX_SHORT_WINDOWS; i++)
ltpInfo->sbk_prediction_used[i] = ltpInfo->delay[i] = 0;
for(i = 0; i < MAX_SCFAC_BANDS; i++)
ltpInfo->sfb_prediction_used[i] = 0;
ltpInfo->side_info = LEN_LTP_DATA_PRESENT;
for(i = 0; i < 2 * BLOCK_LEN_LONG; i++)
ltpInfo->mdct_predicted[i] = 0.0;
}
}
void LtpEnd(faacEncHandle hEncoder)
{
unsigned int channel;
for (channel = 0; channel < hEncoder->numChannels; channel++) {
LtpInfo *ltpInfo = &(hEncoder->coderInfo[channel].ltpInfo);
if (ltpInfo->buffer) FreeMemory(ltpInfo->buffer);
if (ltpInfo->mdct_predicted) FreeMemory(ltpInfo->mdct_predicted);
}
}
int LtpEncode(faacEncHandle hEncoder,
CoderInfo *coderInfo,
LtpInfo *ltpInfo,
TnsInfo *tnsInfo,
double *p_spectrum,
double *p_time_signal)
{
int i, last_band;
double num_bit[MAX_SHORT_WINDOWS];
double *predicted_samples;
ltpInfo->global_pred_flag = 0;
ltpInfo->side_info = 0;
predicted_samples = (double*)AllocMemory(2*BLOCK_LEN_LONG*sizeof(double));
switch(coderInfo->block_type)
{
case ONLY_LONG_WINDOW:
case LONG_SHORT_WINDOW:
case SHORT_LONG_WINDOW:
last_band = (coderInfo->nr_of_sfb < MAX_LT_PRED_LONG_SFB) ? coderInfo->nr_of_sfb : MAX_LT_PRED_LONG_SFB;
ltpInfo->delay[0] =
pitch(p_time_signal, ltpInfo->buffer, 2 * BLOCK_LEN_LONG,
0, 2 * BLOCK_LEN_LONG, predicted_samples, <pInfo->weight,
<pInfo->weight_idx);
num_bit[0] =
ltp_enc_tf(hEncoder, coderInfo, p_spectrum, predicted_samples,
ltpInfo->mdct_predicted,
coderInfo->sfb_offset, coderInfo->nr_of_sfb,
last_band, ltpInfo->side_info, ltpInfo->sfb_prediction_used,
tnsInfo);
ltpInfo->global_pred_flag = (num_bit[0] == 0.0) ? 0 : 1;
if(ltpInfo->global_pred_flag)
for (i = 0; i < coderInfo->sfb_offset[last_band]; i++)
p_spectrum[i] -= ltpInfo->mdct_predicted[i];
else
ltpInfo->side_info = 1;
break;
default:
break;
}
if (predicted_samples) FreeMemory(predicted_samples);
return (ltpInfo->global_pred_flag);
}
void LtpReconstruct(CoderInfo *coderInfo, LtpInfo *ltpInfo, double *p_spectrum)
{
int i, last_band;
if(ltpInfo->global_pred_flag)
{
switch(coderInfo->block_type)
{
case ONLY_LONG_WINDOW:
case LONG_SHORT_WINDOW:
case SHORT_LONG_WINDOW:
last_band = (coderInfo->nr_of_sfb < MAX_LT_PRED_LONG_SFB) ?
coderInfo->nr_of_sfb : MAX_LT_PRED_LONG_SFB;
for (i = 0; i < coderInfo->sfb_offset[last_band]; i++)
p_spectrum[i] += ltpInfo->mdct_predicted[i];
break;
default:
break;
}
}
}
void LtpUpdate(LtpInfo *ltpInfo, double *time_signal,
double *overlap_signal, int block_size_long)
{
int i;
for(i = 0; i < NOK_LT_BLEN - 2 * block_size_long; i++)
ltpInfo->buffer[i] = ltpInfo->buffer[i + block_size_long];
for(i = 0; i < block_size_long; i++)
{
ltpInfo->buffer[NOK_LT_BLEN - 2 * block_size_long + i] = time_signal[i];
ltpInfo->buffer[NOK_LT_BLEN - block_size_long + i] = overlap_signal[i];
}
}