1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
|
/*
* AAC encoder TNS
* Copyright (C) 2015 Rostislav Pehlivanov
*
* 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
*/
/**
* @file
* AAC encoder temporal noise shaping
* @author Rostislav Pehlivanov ( atomnuker gmail com )
*/
#include "aacenc.h"
#include "aacenc_tns.h"
#include "aactab.h"
#include "aacenc_utils.h"
#include "aacenc_quantization.h"
/*
* Shifts the values as well if compression is possible.
*/
static inline int compress_coeffs(int *coef, int order, int c_bits)
{
int i, res = 0;
const int low_idx = c_bits ? 4 : 2;
const int shift_val = c_bits ? 8 : 4;
const int high_idx = c_bits ? 11 : 5;
for (i = 0; i < order; i++)
if (coef[i] < low_idx || coef[i] > high_idx)
res++;
if (res == order)
for (i = 0; i < order; i++)
coef[i] -= (coef[i] > high_idx) ? shift_val : 0;
return res == order;
}
/**
* Encode TNS data.
* Coefficient compression saves a single bit per coefficient.
*/
void ff_aac_encode_tns_info(AACEncContext *s, SingleChannelElement *sce)
{
int i, w, filt, coef_len, coef_compress = 0;
const int is8 = sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE;
TemporalNoiseShaping *tns = &sce->tns;
const int c_bits = is8 ? TNS_Q_BITS_SHORT == 4 : TNS_Q_BITS == 4;
if (!sce->tns.present)
return;
for (i = 0; i < sce->ics.num_windows; i++) {
put_bits(&s->pb, 2 - is8, sce->tns.n_filt[i]);
if (tns->n_filt[i]) {
put_bits(&s->pb, 1, c_bits);
for (filt = 0; filt < tns->n_filt[i]; filt++) {
put_bits(&s->pb, 6 - 2 * is8, tns->length[i][filt]);
put_bits(&s->pb, 5 - 2 * is8, tns->order[i][filt]);
if (tns->order[i][filt]) {
coef_compress = compress_coeffs(tns->coef_idx[i][filt],
tns->order[i][filt], c_bits);
put_bits(&s->pb, 1, !!tns->direction[i][filt]);
put_bits(&s->pb, 1, !!coef_compress);
coef_len = c_bits + 3 - coef_compress;
for (w = 0; w < tns->order[i][filt]; w++)
put_bits(&s->pb, coef_len, tns->coef_idx[i][filt][w]);
}
}
}
}
}
/* Apply TNS filter */
void ff_aac_apply_tns(AACEncContext *s, SingleChannelElement *sce)
{
TemporalNoiseShaping *tns = &sce->tns;
IndividualChannelStream *ics = &sce->ics;
int w, filt, m, i, top, order, bottom, start, end, size, inc;
const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
float lpc[TNS_MAX_ORDER];
for (w = 0; w < ics->num_windows; w++) {
bottom = ics->num_swb;
for (filt = 0; filt < tns->n_filt[w]; filt++) {
top = bottom;
bottom = FFMAX(0, top - tns->length[w][filt]);
order = tns->order[w][filt];
if (order == 0)
continue;
// tns_decode_coef
compute_lpc_coefs(tns->coef[w][filt], order, lpc, 0, 0, 0);
start = ics->swb_offset[FFMIN(bottom, mmm)];
end = ics->swb_offset[FFMIN( top, mmm)];
if ((size = end - start) <= 0)
continue;
if (tns->direction[w][filt]) {
inc = -1;
start = end - 1;
} else {
inc = 1;
}
start += w * 128;
// ar filter
for (m = 0; m < size; m++, start += inc)
for (i = 1; i <= FFMIN(m, order); i++)
sce->coeffs[start] += lpc[i-1]*sce->pcoeffs[start - i*inc];
}
}
}
/*
* c_bits - 1 if 4 bit coefficients, 0 if 3 bit coefficients
*/
static inline void quantize_coefs(double *coef, int *idx, float *lpc, int order,
int c_bits)
{
int i;
const float *quant_arr = tns_tmp2_map[c_bits];
for (i = 0; i < order; i++) {
idx[i] = quant_array_idx((float)coef[i], quant_arr, c_bits ? 16 : 8);
lpc[i] = quant_arr[idx[i]];
}
}
/*
* 3 bits per coefficient with 8 short windows
*/
void ff_aac_search_for_tns(AACEncContext *s, SingleChannelElement *sce)
{
TemporalNoiseShaping *tns = &sce->tns;
int w, w2, g, count = 0;
const int mmm = FFMIN(sce->ics.tns_max_bands, sce->ics.max_sfb);
const int is8 = sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE;
const int c_bits = is8 ? TNS_Q_BITS_SHORT == 4 : TNS_Q_BITS == 4;
int sfb_start = av_clip(tns_min_sfb[is8][s->samplerate_index], 0, mmm);
int sfb_end = av_clip(sce->ics.num_swb, 0, mmm);
for (w = 0; w < sce->ics.num_windows; w++) {
int use_tns;
int order = is8 ? 5 : s->profile == FF_PROFILE_AAC_LOW ? 12 : TNS_MAX_ORDER;
int coef_start = w*sce->ics.num_swb + sce->ics.swb_offset[sfb_start];
int coef_len = sce->ics.swb_offset[sfb_end] - sce->ics.swb_offset[sfb_start];
float e_ratio = 0.0f, threshold = 0.0f, spread = 0.0f, en[2] = {0.0, 0.0f};
double gain = 0.0f, coefs[MAX_LPC_ORDER] = {0};
for (g = 0; g < sce->ics.num_swb; g++) {
if (w*16+g < sfb_start || w*16+g > sfb_end)
continue;
for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
if ((w+w2)*16+g > sfb_start + ((sfb_end - sfb_start)/2))
en[1] += band->energy;
else
en[0] += band->energy;
threshold += band->threshold;
spread += band->spread;
}
}
if (coef_len <= 0 || (sfb_end - sfb_start) <= 0)
continue;
/* LPC */
gain = ff_lpc_calc_ref_coefs_f(&s->lpc, &sce->coeffs[coef_start],
coef_len, order, coefs);
if (!order || gain < TNS_GAIN_THRESHOLD_LOW || gain > TNS_GAIN_THRESHOLD_HIGH)
use_tns = 0;
else if ((en[0]+en[1]) < TNS_GAIN_THRESHOLD_LOW*threshold || spread < TNS_SPREAD_THRESHOLD)
use_tns = 0;
else
use_tns = 1;
if (use_tns) {
e_ratio = en[0]/en[1];
if (is8 || order < 2 || (e_ratio > TNS_E_RATIO_LOW && e_ratio < TNS_E_RATIO_HIGH)) {
tns->n_filt[w] = 1;
for (g = 0; g < tns->n_filt[w]; g++) {
tns->length[w][g] = sfb_end - sfb_start;
tns->direction[w][g] = en[0] < en[1];
tns->order[w][g] = order;
quantize_coefs(coefs, tns->coef_idx[w][g], tns->coef[w][g],
order, c_bits);
}
} else { /* 2 filters due to energy disbalance */
tns->n_filt[w] = 2;
for (g = 0; g < tns->n_filt[w]; g++) {
tns->direction[w][g] = en[g] < en[!g];
tns->order[w][g] = !g ? order/2 : order - tns->order[w][g-1];
tns->length[w][g] = !g ? (sfb_end - sfb_start)/2 : \
(sfb_end - sfb_start) - tns->length[w][g-1];
quantize_coefs(&coefs[!g ? 0 : order - tns->order[w][g-1]],
tns->coef_idx[w][g], tns->coef[w][g],
tns->order[w][g], c_bits);
}
}
count++;
}
}
sce->tns.present = !!count;
}
|
