cb377ec5 |
/*
* QCELP decoder
* Copyright (c) 2007 Reynaldo H. Verdejo Pinochet
*
* This file is part of FFmpeg.
*
* FFmpeg 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.
*
* FFmpeg 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 FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/ |
1c3ae1ab |
|
cb377ec5 |
/** |
ba87f080 |
* @file |
cb377ec5 |
* QCELP decoder
* @author Reynaldo H. Verdejo Pinochet |
8e36385a |
* @remark FFmpeg merging spearheaded by Kenan Gillet |
9e00c20e |
* @remark Development mentored by Benjamin Larson |
cb377ec5 |
*/
#include <stddef.h>
|
1fa0284c |
#include "libavutil/avassert.h" |
a903f8f0 |
#include "libavutil/channel_layout.h" |
d56668bd |
#include "libavutil/float_dsp.h" |
cb377ec5 |
#include "avcodec.h" |
dbbec0c2 |
#include "internal.h" |
9106a698 |
#include "get_bits.h" |
cb377ec5 |
#include "qcelpdata.h"
#include "celp_filters.h" |
805a83d5 |
#include "acelp_filters.h" |
95e83257 |
#include "acelp_vectors.h" |
33ae681f |
#include "lsp.h" |
cb377ec5 |
|
25c2d76b |
typedef enum { |
adbfc605 |
I_F_Q = -1, /**< insufficient frame quality */ |
061f407e |
SILENCE,
RATE_OCTAVE,
RATE_QUARTER,
RATE_HALF,
RATE_FULL
} qcelp_packet_rate;
|
7f9f771e |
typedef struct QCELPContext { |
640760da |
GetBitContext gb;
qcelp_packet_rate bitrate; |
adbfc605 |
QCELPFrame frame; /**< unpacked data frame */ |
cf139541 |
uint8_t erasure_count; |
adbfc605 |
uint8_t octave_count; /**< count the consecutive RATE_OCTAVE frames */ |
cf139541 |
float prev_lspf[10]; |
adbfc605 |
float predictor_lspf[10];/**< LSP predictor for RATE_OCTAVE and I_F_Q */ |
cf139541 |
float pitch_synthesis_filter_mem[303];
float pitch_pre_filter_mem[303];
float rnd_fir_filter_mem[180];
float formant_mem[170];
float last_codebook_gain;
int prev_g1[2];
int prev_bitrate;
float pitch_gain[4];
uint8_t pitch_lag[4];
uint16_t first16bits; |
5a3e9f2c |
uint8_t warned_buf_mismatch_bitrate; |
805a83d5 |
/* postfilter */
float postfilter_synth_mem[10];
float postfilter_agc_mem;
float postfilter_tilt_mem; |
640760da |
} QCELPContext;
|
061f407e |
/** |
39ded680 |
* Initialize the speech codec according to the specification.
*
* TIA/EIA/IS-733 2.4.9
*/ |
1b321c5c |
static av_cold int qcelp_decode_init(AVCodecContext *avctx)
{ |
39ded680 |
QCELPContext *q = avctx->priv_data;
int i;
|
e3d6ab57 |
avctx->channels = 1;
avctx->channel_layout = AV_CH_LAYOUT_MONO;
avctx->sample_fmt = AV_SAMPLE_FMT_FLT; |
39ded680 |
|
25c2d76b |
for (i = 0; i < 10; i++) |
4a2ef394 |
q->prev_lspf[i] = (i + 1) / 11.0; |
39ded680 |
return 0;
}
/** |
49bd8e4b |
* Decode the 10 quantized LSP frequencies from the LSPV/LSP
* transmission codes of any bitrate and check for badly received packets. |
148c31b9 |
*
* @param q the context
* @param lspf line spectral pair frequencies
*
* @return 0 on success, -1 if the packet is badly received
*
* TIA/EIA/IS-733 2.4.3.2.6.2-2, 2.4.8.7.3
*/ |
1b321c5c |
static int decode_lspf(QCELPContext *q, float *lspf)
{ |
148c31b9 |
int i; |
cf139541 |
float tmp_lspf, smooth, erasure_coeff;
const float *predictors; |
148c31b9 |
|
dd376b1a |
if (q->bitrate == RATE_OCTAVE || q->bitrate == I_F_Q) { |
2fd29184 |
predictors = q->prev_bitrate != RATE_OCTAVE &&
q->prev_bitrate != I_F_Q ? q->prev_lspf
: q->predictor_lspf; |
148c31b9 |
|
dd376b1a |
if (q->bitrate == RATE_OCTAVE) { |
148c31b9 |
q->octave_count++;
|
25c2d76b |
for (i = 0; i < 10; i++) { |
148c31b9 |
q->predictor_lspf[i] = |
640760da |
lspf[i] = (q->frame.lspv[i] ? QCELP_LSP_SPREAD_FACTOR |
25c2d76b |
: -QCELP_LSP_SPREAD_FACTOR) +
predictors[i] * QCELP_LSP_OCTAVE_PREDICTOR +
(i + 1) * ((1 - QCELP_LSP_OCTAVE_PREDICTOR) / 11); |
148c31b9 |
} |
2fd29184 |
smooth = q->octave_count < 10 ? .875 : 0.1; |
dd376b1a |
} else { |
cf139541 |
erasure_coeff = QCELP_LSP_OCTAVE_PREDICTOR; |
148c31b9 |
|
1fa0284c |
av_assert2(q->bitrate == I_F_Q); |
148c31b9 |
|
25c2d76b |
if (q->erasure_count > 1) |
2fd29184 |
erasure_coeff *= q->erasure_count < 4 ? 0.9 : 0.7; |
148c31b9 |
|
25c2d76b |
for (i = 0; i < 10; i++) { |
148c31b9 |
q->predictor_lspf[i] = |
25c2d76b |
lspf[i] = (i + 1) * (1 - erasure_coeff) / 11 +
erasure_coeff * predictors[i]; |
148c31b9 |
}
smooth = 0.125;
}
// Check the stability of the LSP frequencies.
lspf[0] = FFMAX(lspf[0], QCELP_LSP_SPREAD_FACTOR); |
25c2d76b |
for (i = 1; i < 10; i++) |
2fd29184 |
lspf[i] = FFMAX(lspf[i], lspf[i - 1] + QCELP_LSP_SPREAD_FACTOR); |
148c31b9 |
|
2fd29184 |
lspf[9] = FFMIN(lspf[9], 1.0 - QCELP_LSP_SPREAD_FACTOR); |
25c2d76b |
for (i = 9; i > 0; i--) |
2fd29184 |
lspf[i - 1] = FFMIN(lspf[i - 1], lspf[i] - QCELP_LSP_SPREAD_FACTOR); |
148c31b9 |
// Low-pass filter the LSP frequencies. |
25c2d76b |
ff_weighted_vector_sumf(lspf, lspf, q->prev_lspf, smooth, 1.0 - smooth, 10); |
dd376b1a |
} else { |
148c31b9 |
q->octave_count = 0;
|
4a2ef394 |
tmp_lspf = 0.0; |
dd376b1a |
for (i = 0; i < 5; i++) { |
25c2d76b |
lspf[2 * i + 0] = tmp_lspf += qcelp_lspvq[i][q->frame.lspv[i]][0] * 0.0001;
lspf[2 * i + 1] = tmp_lspf += qcelp_lspvq[i][q->frame.lspv[i]][1] * 0.0001; |
148c31b9 |
}
// Check for badly received packets. |
dd376b1a |
if (q->bitrate == RATE_QUARTER) { |
25c2d76b |
if (lspf[9] <= .70 || lspf[9] >= .97) |
148c31b9 |
return -1; |
25c2d76b |
for (i = 3; i < 10; i++)
if (fabs(lspf[i] - lspf[i - 2]) < .08) |
148c31b9 |
return -1; |
dd376b1a |
} else { |
25c2d76b |
if (lspf[9] <= .66 || lspf[9] >= .985) |
148c31b9 |
return -1; |
25c2d76b |
for (i = 4; i < 10; i++)
if (fabs(lspf[i] - lspf[i - 4]) < .0931) |
148c31b9 |
return -1;
}
}
return 0;
}
/** |
49bd8e4b |
* Convert codebook transmission codes to GAIN and INDEX. |
640760da |
*
* @param q the context
* @param gain array holding the decoded gain
*
* TIA/EIA/IS-733 2.4.6.2
*/ |
25c2d76b |
static void decode_gain_and_index(QCELPContext *q, float *gain)
{
int i, subframes_count, g1[16]; |
640760da |
float slope;
|
dd376b1a |
if (q->bitrate >= RATE_QUARTER) {
switch (q->bitrate) { |
25c2d76b |
case RATE_FULL: subframes_count = 16; break;
case RATE_HALF: subframes_count = 4; break;
default: subframes_count = 5; |
640760da |
} |
25c2d76b |
for (i = 0; i < subframes_count; i++) { |
640760da |
g1[i] = 4 * q->frame.cbgain[i]; |
25c2d76b |
if (q->bitrate == RATE_FULL && !((i + 1) & 3)) {
g1[i] += av_clip((g1[i - 1] + g1[i - 2] + g1[i - 3]) / 3 - 6, 0, 32); |
640760da |
}
gain[i] = qcelp_g12ga[g1[i]];
|
dd376b1a |
if (q->frame.cbsign[i]) { |
640760da |
gain[i] = -gain[i]; |
25c2d76b |
q->frame.cindex[i] = (q->frame.cindex[i] - 89) & 127; |
640760da |
}
}
|
25c2d76b |
q->prev_g1[0] = g1[i - 2];
q->prev_g1[1] = g1[i - 1];
q->last_codebook_gain = qcelp_g12ga[g1[i - 1]]; |
640760da |
|
dd376b1a |
if (q->bitrate == RATE_QUARTER) { |
640760da |
// Provide smoothing of the unvoiced excitation energy. |
25c2d76b |
gain[7] = gain[4];
gain[6] = 0.4 * gain[3] + 0.6 * gain[4];
gain[5] = gain[3];
gain[4] = 0.8 * gain[2] + 0.2 * gain[3];
gain[3] = 0.2 * gain[1] + 0.8 * gain[2];
gain[2] = gain[1];
gain[1] = 0.6 * gain[0] + 0.4 * gain[1]; |
640760da |
} |
dd376b1a |
} else if (q->bitrate != SILENCE) {
if (q->bitrate == RATE_OCTAVE) { |
25c2d76b |
g1[0] = 2 * q->frame.cbgain[0] +
av_clip((q->prev_g1[0] + q->prev_g1[1]) / 2 - 5, 0, 54); |
640760da |
subframes_count = 8; |
dd376b1a |
} else { |
1fa0284c |
av_assert2(q->bitrate == I_F_Q); |
640760da |
g1[0] = q->prev_g1[1]; |
dd376b1a |
switch (q->erasure_count) { |
25c2d76b |
case 1 : break;
case 2 : g1[0] -= 1; break;
case 3 : g1[0] -= 2; break;
default: g1[0] -= 6; |
640760da |
} |
25c2d76b |
if (g1[0] < 0) |
640760da |
g1[0] = 0;
subframes_count = 4;
}
// This interpolation is done to produce smoother background noise. |
25c2d76b |
slope = 0.5 * (qcelp_g12ga[g1[0]] - q->last_codebook_gain) / subframes_count;
for (i = 1; i <= subframes_count; i++)
gain[i - 1] = q->last_codebook_gain + slope * i; |
640760da |
|
25c2d76b |
q->last_codebook_gain = gain[i - 2];
q->prev_g1[0] = q->prev_g1[1];
q->prev_g1[1] = g1[0]; |
640760da |
}
}
/** |
1b321c5c |
* If the received packet is Rate 1/4 a further sanity check is made of the
* codebook gain. |
148c31b9 |
*
* @param cbgain the unpacked cbgain array
* @return -1 if the sanity check fails, 0 otherwise
*
* TIA/EIA/IS-733 2.4.8.7.3
*/ |
1b321c5c |
static int codebook_sanity_check_for_rate_quarter(const uint8_t *cbgain)
{ |
25c2d76b |
int i, diff, prev_diff = 0; |
148c31b9 |
|
25c2d76b |
for (i = 1; i < 5; i++) { |
cf139541 |
diff = cbgain[i] - cbgain[i-1]; |
25c2d76b |
if (FFABS(diff) > 10) |
1b321c5c |
return -1; |
25c2d76b |
else if (FFABS(diff - prev_diff) > 12) |
1b321c5c |
return -1;
prev_diff = diff; |
8372e3d2 |
}
return 0; |
148c31b9 |
}
/** |
49bd8e4b |
* Compute the scaled codebook vector Cdn From INDEX and GAIN |
39ded680 |
* for all rates.
*
* The specification lacks some information here.
*
* TIA/EIA/IS-733 has an omission on the codebook index determination
* formula for RATE_FULL and RATE_HALF frames at section 2.4.8.1.1. It says
* you have to subtract the decoded index parameter from the given scaled
* codebook vector index 'n' to get the desired circular codebook index, but
* it does not mention that you have to clamp 'n' to [0-9] in order to get
* RI-compliant results.
*
* The reason for this mistake seems to be the fact they forgot to mention you
* have to do these calculations per codebook subframe and adjust given
* equation values accordingly.
*
* @param q the context
* @param gain array holding the 4 pitch subframe gain values
* @param cdn_vector array for the generated scaled codebook vector
*/ |
b12c7627 |
static void compute_svector(QCELPContext *q, const float *gain, |
1b321c5c |
float *cdn_vector)
{ |
25c2d76b |
int i, j, k; |
39ded680 |
uint16_t cbseed, cindex; |
25c2d76b |
float *rnd, tmp_gain, fir_filter_value; |
39ded680 |
|
dd376b1a |
switch (q->bitrate) { |
25c2d76b |
case RATE_FULL:
for (i = 0; i < 16; i++) {
tmp_gain = gain[i] * QCELP_RATE_FULL_CODEBOOK_RATIO;
cindex = -q->frame.cindex[i];
for (j = 0; j < 10; j++) |
84cdd2fd |
*cdn_vector++ = tmp_gain *
qcelp_rate_full_codebook[cindex++ & 127]; |
25c2d76b |
} |
39ded680 |
break; |
25c2d76b |
case RATE_HALF:
for (i = 0; i < 4; i++) {
tmp_gain = gain[i] * QCELP_RATE_HALF_CODEBOOK_RATIO;
cindex = -q->frame.cindex[i];
for (j = 0; j < 40; j++) |
84cdd2fd |
*cdn_vector++ = tmp_gain *
qcelp_rate_half_codebook[cindex++ & 127]; |
25c2d76b |
} |
39ded680 |
break; |
25c2d76b |
case RATE_QUARTER:
cbseed = (0x0003 & q->frame.lspv[4]) << 14 |
(0x003F & q->frame.lspv[3]) << 8 |
(0x0060 & q->frame.lspv[2]) << 1 |
(0x0007 & q->frame.lspv[1]) << 3 |
(0x0038 & q->frame.lspv[0]) >> 3;
rnd = q->rnd_fir_filter_mem + 20;
for (i = 0; i < 8; i++) {
tmp_gain = gain[i] * (QCELP_SQRT1887 / 32768.0);
for (k = 0; k < 20; k++) {
cbseed = 521 * cbseed + 259;
*rnd = (int16_t) cbseed; |
1b321c5c |
// FIR filter |
25c2d76b |
fir_filter_value = 0.0;
for (j = 0; j < 10; j++)
fir_filter_value += qcelp_rnd_fir_coefs[j] *
(rnd[-j] + rnd[-20+j]);
fir_filter_value += qcelp_rnd_fir_coefs[10] * rnd[-10];
*cdn_vector++ = tmp_gain * fir_filter_value;
rnd++; |
39ded680 |
} |
25c2d76b |
}
memcpy(q->rnd_fir_filter_mem, q->rnd_fir_filter_mem + 160,
20 * sizeof(float)); |
39ded680 |
break; |
25c2d76b |
case RATE_OCTAVE:
cbseed = q->first16bits;
for (i = 0; i < 8; i++) {
tmp_gain = gain[i] * (QCELP_SQRT1887 / 32768.0);
for (j = 0; j < 20; j++) {
cbseed = 521 * cbseed + 259;
*cdn_vector++ = tmp_gain * (int16_t) cbseed; |
39ded680 |
} |
25c2d76b |
} |
39ded680 |
break; |
25c2d76b |
case I_F_Q:
cbseed = -44; // random codebook index
for (i = 0; i < 4; i++) {
tmp_gain = gain[i] * QCELP_RATE_FULL_CODEBOOK_RATIO;
for (j = 0; j < 40; j++) |
84cdd2fd |
*cdn_vector++ = tmp_gain *
qcelp_rate_full_codebook[cbseed++ & 127]; |
25c2d76b |
} |
39ded680 |
break; |
25c2d76b |
case SILENCE:
memset(cdn_vector, 0, 160 * sizeof(float)); |
4f54fb54 |
break; |
39ded680 |
}
}
/** |
809e07c6 |
* Apply generic gain control.
*
* @param v_out output vector
* @param v_in gain-controlled vector
* @param v_ref vector to control gain of
*
* TIA/EIA/IS-733 2.4.8.3, 2.4.8.6
*/ |
25c2d76b |
static void apply_gain_ctrl(float *v_out, const float *v_ref, const float *v_in) |
1b321c5c |
{ |
0c50f8e6 |
int i; |
39ded680 |
|
d56668bd |
for (i = 0; i < 160; i += 40) {
float res = avpriv_scalarproduct_float_c(v_ref + i, v_ref + i, 40);
ff_scale_vector_to_given_sum_of_squares(v_out + i, v_in + i, res, 40);
} |
39ded680 |
}
/** |
cb377ec5 |
* Apply filter in pitch-subframe steps.
*
* @param memory buffer for the previous state of the filter
* - must be able to contain 303 elements
* - the 143 first elements are from the previous state
* - the next 160 are for output
* @param v_in input filter vector
* @param gain per-subframe gain array, each element is between 0.0 and 2.0
* @param lag per-subframe lag array, each element is
* - between 16 and 143 if its corresponding pfrac is 0,
* - between 16 and 139 otherwise |
1b321c5c |
* @param pfrac per-subframe boolean array, 1 if the lag is fractional, 0
* otherwise |
cb377ec5 |
*
* @return filter output vector
*/ |
1c3ae1ab |
static const float *do_pitchfilter(float memory[303], const float v_in[160],
const float gain[4], const uint8_t *lag,
const uint8_t pfrac[4])
{ |
25c2d76b |
int i, j;
float *v_lag, *v_out; |
cb377ec5 |
const float *v_len;
v_out = memory + 143; // Output vector starts at memory[143].
|
dd376b1a |
for (i = 0; i < 4; i++) {
if (gain[i]) { |
cb377ec5 |
v_lag = memory + 143 + 40 * i - lag[i]; |
dd376b1a |
for (v_len = v_in + 40; v_in < v_len; v_in++) {
if (pfrac[i]) { // If it is a fractional lag... |
4a2ef394 |
for (j = 0, *v_out = 0.0; j < 4; j++) |
84cdd2fd |
*v_out += qcelp_hammsinc_table[j] *
(v_lag[j - 4] + v_lag[3 - j]); |
25c2d76b |
} else |
cb377ec5 |
*v_out = *v_lag;
*v_out = *v_in + gain[i] * *v_out;
v_lag++;
v_out++;
} |
dd376b1a |
} else { |
cb377ec5 |
memcpy(v_out, v_in, 40 * sizeof(float));
v_in += 40;
v_out += 40;
} |
1c3ae1ab |
} |
cb377ec5 |
memmove(memory, memory + 160, 143 * sizeof(float));
return memory + 143;
}
|
200de8c6 |
/** |
73b458e3 |
* Apply pitch synthesis filter and pitch prefilter to the scaled codebook vector. |
4f54fb54 |
* TIA/EIA/IS-733 2.4.5.2, 2.4.8.7.2 |
73b458e3 |
*
* @param q the context
* @param cdn_vector the scaled codebook vector
*/ |
cf139541 |
static void apply_pitch_filters(QCELPContext *q, float *cdn_vector)
{ |
25c2d76b |
int i; |
73b458e3 |
const float *v_synthesis_filtered, *v_pre_filtered;
|
25c2d76b |
if (q->bitrate >= RATE_HALF || q->bitrate == SILENCE ||
(q->bitrate == I_F_Q && (q->prev_bitrate >= RATE_HALF))) { |
73b458e3 |
|
25c2d76b |
if (q->bitrate >= RATE_HALF) { |
73b458e3 |
// Compute gain & lag for the whole frame. |
dd376b1a |
for (i = 0; i < 4; i++) { |
73b458e3 |
q->pitch_gain[i] = q->frame.plag[i] ? (q->frame.pgain[i] + 1) * 0.25 : 0.0;
q->pitch_lag[i] = q->frame.plag[i] + 16;
} |
dd376b1a |
} else { |
3f16ed15 |
float max_pitch_gain;
|
dd376b1a |
if (q->bitrate == I_F_Q) { |
a3636fa0 |
if (q->erasure_count < 3)
max_pitch_gain = 0.9 - 0.3 * (q->erasure_count - 1);
else
max_pitch_gain = 0.0; |
dd376b1a |
} else { |
1fa0284c |
av_assert2(q->bitrate == SILENCE); |
4f54fb54 |
max_pitch_gain = 1.0;
} |
25c2d76b |
for (i = 0; i < 4; i++) |
73b458e3 |
q->pitch_gain[i] = FFMIN(q->pitch_gain[i], max_pitch_gain);
memset(q->frame.pfrac, 0, sizeof(q->frame.pfrac));
}
// pitch synthesis filter |
cf139541 |
v_synthesis_filtered = do_pitchfilter(q->pitch_synthesis_filter_mem,
cdn_vector, q->pitch_gain,
q->pitch_lag, q->frame.pfrac); |
73b458e3 |
// pitch prefilter update |
25c2d76b |
for (i = 0; i < 4; i++) |
73b458e3 |
q->pitch_gain[i] = 0.5 * FFMIN(q->pitch_gain[i], 1.0);
|
25c2d76b |
v_pre_filtered = do_pitchfilter(q->pitch_pre_filter_mem,
v_synthesis_filtered,
q->pitch_gain, q->pitch_lag,
q->frame.pfrac); |
73b458e3 |
apply_gain_ctrl(cdn_vector, v_synthesis_filtered, v_pre_filtered); |
dd376b1a |
} else { |
84cdd2fd |
memcpy(q->pitch_synthesis_filter_mem,
cdn_vector + 17, 143 * sizeof(float)); |
cf139541 |
memcpy(q->pitch_pre_filter_mem, cdn_vector + 17, 143 * sizeof(float)); |
73b458e3 |
memset(q->pitch_gain, 0, sizeof(q->pitch_gain));
memset(q->pitch_lag, 0, sizeof(q->pitch_lag));
}
}
/** |
49bd8e4b |
* Reconstruct LPC coefficients from the line spectral pair frequencies
* and perform bandwidth expansion. |
87574416 |
*
* @param lspf line spectral pair frequencies
* @param lpc linear predictive coding coefficients
* |
e19f9952 |
* @note: bandwidth_expansion_coeff could be precalculated into a table |
87574416 |
* but it seems to be slower on x86
*
* TIA/EIA/IS-733 2.4.3.3.5
*/ |
3c00556f |
static void lspf2lpc(const float *lspf, float *lpc) |
87574416 |
{ |
1e1e02ea |
double lsp[10]; |
e19f9952 |
double bandwidth_expansion_coeff = QCELP_BANDWIDTH_EXPANSION_COEFF; |
25c2d76b |
int i; |
87574416 |
|
25c2d76b |
for (i = 0; i < 10; i++) |
1e1e02ea |
lsp[i] = cos(M_PI * lspf[i]); |
87574416 |
|
e26d1318 |
ff_acelp_lspd2lpc(lsp, lpc, 5); |
87574416 |
|
dd376b1a |
for (i = 0; i < 10; i++) { |
25c2d76b |
lpc[i] *= bandwidth_expansion_coeff; |
e19f9952 |
bandwidth_expansion_coeff *= QCELP_BANDWIDTH_EXPANSION_COEFF; |
87574416 |
}
}
/** |
48966b02 |
* Interpolate LSP frequencies and compute LPC coefficients |
148c31b9 |
* for a given bitrate & pitch subframe. |
200de8c6 |
* |
4f54fb54 |
* TIA/EIA/IS-733 2.4.3.3.4, 2.4.8.7.2 |
200de8c6 |
*
* @param q the context
* @param curr_lspf LSP frequencies vector of the current frame
* @param lpc float vector for the resulting LPC
* @param subframe_num frame number in decoded stream
*/ |
0e426b7a |
static void interpolate_lpc(QCELPContext *q, const float *curr_lspf,
float *lpc, const int subframe_num) |
1c3ae1ab |
{ |
200de8c6 |
float interpolated_lspf[10];
float weight;
|
25c2d76b |
if (q->bitrate >= RATE_QUARTER) |
200de8c6 |
weight = 0.25 * (subframe_num + 1); |
25c2d76b |
else if (q->bitrate == RATE_OCTAVE && !subframe_num) |
200de8c6 |
weight = 0.625; |
1c3ae1ab |
else |
200de8c6 |
weight = 1.0;
|
dd376b1a |
if (weight != 1.0) { |
95e83257 |
ff_weighted_vector_sumf(interpolated_lspf, curr_lspf, q->prev_lspf, |
d2af5697 |
weight, 1.0 - weight, 10); |
87574416 |
lspf2lpc(interpolated_lspf, lpc); |
25c2d76b |
} else if (q->bitrate >= RATE_QUARTER ||
(q->bitrate == I_F_Q && !subframe_num)) |
87574416 |
lspf2lpc(curr_lspf, lpc); |
25c2d76b |
else if (q->bitrate == SILENCE && !subframe_num) |
87574416 |
lspf2lpc(q->prev_lspf, lpc); |
200de8c6 |
}
|
6e74619e |
static qcelp_packet_rate buf_size2bitrate(const int buf_size) |
1c3ae1ab |
{ |
dd376b1a |
switch (buf_size) { |
25c2d76b |
case 35: return RATE_FULL;
case 17: return RATE_HALF;
case 8: return RATE_QUARTER;
case 4: return RATE_OCTAVE;
case 1: return SILENCE; |
2ae1a9b2 |
} |
1c3ae1ab |
|
6e74619e |
return I_F_Q; |
2ae1a9b2 |
}
|
061f407e |
/**
* Determine the bitrate from the frame size and/or the first byte of the frame.
*
* @param avctx the AV codec context
* @param buf_size length of the buffer
* @param buf the bufffer
*
* @return the bitrate on success,
* I_F_Q if the bitrate cannot be satisfactorily determined
*
* TIA/EIA/IS-733 2.4.8.7.1
*/ |
25c2d76b |
static qcelp_packet_rate determine_bitrate(AVCodecContext *avctx,
const int buf_size,
const uint8_t **buf) |
cf139541 |
{ |
061f407e |
qcelp_packet_rate bitrate;
|
dd376b1a |
if ((bitrate = buf_size2bitrate(buf_size)) >= 0) {
if (bitrate > **buf) { |
5a3e9f2c |
QCELPContext *q = avctx->priv_data; |
dd376b1a |
if (!q->warned_buf_mismatch_bitrate) { |
cf139541 |
av_log(avctx, AV_LOG_WARNING,
"Claimed bitrate and buffer size mismatch.\n"); |
5a3e9f2c |
q->warned_buf_mismatch_bitrate = 1;
} |
061f407e |
bitrate = **buf; |
dd376b1a |
} else if (bitrate < **buf) { |
cf139541 |
av_log(avctx, AV_LOG_ERROR,
"Buffer is too small for the claimed bitrate.\n"); |
061f407e |
return I_F_Q;
}
(*buf)++; |
dd376b1a |
} else if ((bitrate = buf_size2bitrate(buf_size + 1)) >= 0) { |
061f407e |
av_log(avctx, AV_LOG_WARNING, |
b295bce1 |
"Bitrate byte missing, guessing bitrate from packet size.\n"); |
25c2d76b |
} else |
061f407e |
return I_F_Q;
|
dd376b1a |
if (bitrate == SILENCE) { |
6d97484d |
// FIXME: Remove this warning when tested with samples.
avpriv_request_sample(avctx, "Blank frame handling"); |
061f407e |
}
return bitrate;
}
|
cb377ec5 |
static void warn_insufficient_frame_quality(AVCodecContext *avctx, |
1c3ae1ab |
const char *message)
{ |
25c2d76b |
av_log(avctx, AV_LOG_WARNING, "Frame #%d, IFQ: %s\n",
avctx->frame_number, message); |
cb377ec5 |
} |
200de8c6 |
|
805a83d5 |
static void postfilter(QCELPContext *q, float *samples, float *lpc)
{
static const float pow_0_775[10] = {
0.775000, 0.600625, 0.465484, 0.360750, 0.279582,
0.216676, 0.167924, 0.130141, 0.100859, 0.078166
}, pow_0_625[10] = {
0.625000, 0.390625, 0.244141, 0.152588, 0.095367,
0.059605, 0.037253, 0.023283, 0.014552, 0.009095
};
float lpc_s[10], lpc_p[10], pole_out[170], zero_out[160];
int n;
for (n = 0; n < 10; n++) {
lpc_s[n] = lpc[n] * pow_0_625[n];
lpc_p[n] = lpc[n] * pow_0_775[n];
}
ff_celp_lp_zero_synthesis_filterf(zero_out, lpc_s,
q->formant_mem + 10, 160, 10); |
25c2d76b |
memcpy(pole_out, q->postfilter_synth_mem, sizeof(float) * 10); |
805a83d5 |
ff_celp_lp_synthesis_filterf(pole_out + 10, lpc_p, zero_out, 160, 10);
memcpy(q->postfilter_synth_mem, pole_out + 160, sizeof(float) * 10);
ff_tilt_compensation(&q->postfilter_tilt_mem, 0.3, pole_out + 10, 160);
ff_adaptive_gain_control(samples, pole_out + 10, |
d56668bd |
avpriv_scalarproduct_float_c(q->formant_mem + 10,
q->formant_mem + 10,
160), |
25c2d76b |
160, 0.9375, &q->postfilter_agc_mem); |
805a83d5 |
}
|
0eea2129 |
static int qcelp_decode_frame(AVCodecContext *avctx, void *data,
int *got_frame_ptr, AVPacket *avpkt) |
cf139541 |
{ |
7a00bbad |
const uint8_t *buf = avpkt->data; |
25c2d76b |
int buf_size = avpkt->size;
QCELPContext *q = avctx->priv_data; |
1b9b6d6e |
AVFrame *frame = data; |
0eea2129 |
float *outbuffer;
int i, ret; |
cf139541 |
float quantized_lspf[10], lpc[10];
float gain[16];
float *formant_mem;
|
0eea2129 |
/* get output buffer */ |
1b9b6d6e |
frame->nb_samples = 160; |
1ec94b0f |
if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) |
0eea2129 |
return ret; |
1b9b6d6e |
outbuffer = (float *)frame->data[0]; |
e43dd3d2 |
|
dd376b1a |
if ((q->bitrate = determine_bitrate(avctx, buf_size, &buf)) == I_F_Q) { |
b295bce1 |
warn_insufficient_frame_quality(avctx, "Bitrate cannot be determined."); |
640760da |
goto erasure;
}
|
25c2d76b |
if (q->bitrate == RATE_OCTAVE &&
(q->first16bits = AV_RB16(buf)) == 0xFFFF) { |
640760da |
warn_insufficient_frame_quality(avctx, "Bitrate is 1/8 and first 16 bits are on.");
goto erasure;
}
|
dd376b1a |
if (q->bitrate > SILENCE) { |
640760da |
const QCELPBitmap *bitmaps = qcelp_unpacking_bitmaps_per_rate[q->bitrate]; |
25c2d76b |
const QCELPBitmap *bitmaps_end = qcelp_unpacking_bitmaps_per_rate[q->bitrate] +
qcelp_unpacking_bitmaps_lengths[q->bitrate];
uint8_t *unpacked_data = (uint8_t *)&q->frame; |
640760da |
|
e9eb94ac |
if ((ret = init_get_bits8(&q->gb, buf, buf_size)) < 0)
return ret; |
640760da |
memset(&q->frame, 0, sizeof(QCELPFrame));
|
25c2d76b |
for (; bitmaps < bitmaps_end; bitmaps++) |
640760da |
unpacked_data[bitmaps->index] |= get_bits(&q->gb, bitmaps->bitlen) << bitmaps->bitpos;
// Check for erasures/blanks on rates 1, 1/4 and 1/8. |
dd376b1a |
if (q->frame.reserved) { |
640760da |
warn_insufficient_frame_quality(avctx, "Wrong data in reserved frame area.");
goto erasure;
} |
25c2d76b |
if (q->bitrate == RATE_QUARTER &&
codebook_sanity_check_for_rate_quarter(q->frame.cbgain)) { |
640760da |
warn_insufficient_frame_quality(avctx, "Codebook gain sanity check failed.");
goto erasure;
}
|
dd376b1a |
if (q->bitrate >= RATE_HALF) {
for (i = 0; i < 4; i++) {
if (q->frame.pfrac[i] && q->frame.plag[i] >= 124) { |
640760da |
warn_insufficient_frame_quality(avctx, "Cannot initialize pitch filter.");
goto erasure;
}
}
}
}
decode_gain_and_index(q, gain);
compute_svector(q, gain, outbuffer);
|
dd376b1a |
if (decode_lspf(q, quantized_lspf) < 0) { |
640760da |
warn_insufficient_frame_quality(avctx, "Badly received packets in frame.");
goto erasure;
}
apply_pitch_filters(q, outbuffer);
|
dd376b1a |
if (q->bitrate == I_F_Q) { |
640760da |
erasure:
q->bitrate = I_F_Q;
q->erasure_count++;
decode_gain_and_index(q, gain);
compute_svector(q, gain, outbuffer);
decode_lspf(q, quantized_lspf);
apply_pitch_filters(q, outbuffer); |
25c2d76b |
} else |
640760da |
q->erasure_count = 0;
formant_mem = q->formant_mem + 10; |
dd376b1a |
for (i = 0; i < 4; i++) { |
640760da |
interpolate_lpc(q, quantized_lspf, lpc, i); |
84cdd2fd |
ff_celp_lp_synthesis_filterf(formant_mem, lpc,
outbuffer + i * 40, 40, 10); |
640760da |
formant_mem += 40;
}
|
805a83d5 |
// postfilter, as per TIA/EIA/IS-733 2.4.8.6
postfilter(q, outbuffer, lpc);
memcpy(q->formant_mem, q->formant_mem + 160, 10 * sizeof(float)); |
640760da |
memcpy(q->prev_lspf, quantized_lspf, sizeof(q->prev_lspf)); |
25c2d76b |
q->prev_bitrate = q->bitrate; |
640760da |
|
1b9b6d6e |
*got_frame_ptr = 1; |
640760da |
|
04c13dca |
return buf_size; |
640760da |
}
|
25c2d76b |
AVCodec ff_qcelp_decoder = {
.name = "qcelp", |
b2bed932 |
.long_name = NULL_IF_CONFIG_SMALL("QCELP / PureVoice"), |
25c2d76b |
.type = AVMEDIA_TYPE_AUDIO, |
36ef5369 |
.id = AV_CODEC_ID_QCELP, |
25c2d76b |
.init = qcelp_decode_init,
.decode = qcelp_decode_frame, |
def97856 |
.capabilities = AV_CODEC_CAP_DR1, |
200de8c6 |
.priv_data_size = sizeof(QCELPContext),
}; |