aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/twinvqdec.c
blob: 9e4587f2bc958648cf0f6711a2bdcf2d195c0e07 (plain) (blame)
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
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
/*
 * TwinVQ decoder
 * Copyright (c) 2009 Vitor Sessak
 *
 * 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
 */

#include <math.h>
#include <stdint.h>

#include "libavutil/channel_layout.h"
#include "avcodec.h"
#include "codec_internal.h"
#include "get_bits.h"
#include "twinvq.h"
#include "metasound_data.h"
#include "twinvq_data.h"

static const TwinVQModeTab mode_08_08 = {
    {
        { 8, bark_tab_s08_64,  10, tab.fcb08s, 1, 5, tab.cb0808s0, tab.cb0808s1, 18 },
        { 2, bark_tab_m08_256, 20, tab.fcb08m, 2, 5, tab.cb0808m0, tab.cb0808m1, 16 },
        { 1, bark_tab_l08_512, 30, tab.fcb08l, 3, 6, tab.cb0808l0, tab.cb0808l1, 17 }
    },
    512, 12, ff_metasound_lsp8, 1, 5, 3, 3, tab.shape08, 8, 28, 20, 6, 40
};

static const TwinVQModeTab mode_11_08 = {
    {
        { 8, bark_tab_s11_64,  10, tab.fcb11s, 1, 5, tab.cb1108s0, tab.cb1108s1, 29 },
        { 2, bark_tab_m11_256, 20, tab.fcb11m, 2, 5, tab.cb1108m0, tab.cb1108m1, 24 },
        { 1, bark_tab_l11_512, 30, tab.fcb11l, 3, 6, tab.cb1108l0, tab.cb1108l1, 27 }
    },
    512, 16, ff_metasound_lsp11, 1, 6, 4, 3, tab.shape11, 9, 36, 30, 7, 90
};

static const TwinVQModeTab mode_11_10 = {
    {
        { 8, bark_tab_s11_64,  10, tab.fcb11s, 1, 5, tab.cb1110s0, tab.cb1110s1, 21 },
        { 2, bark_tab_m11_256, 20, tab.fcb11m, 2, 5, tab.cb1110m0, tab.cb1110m1, 18 },
        { 1, bark_tab_l11_512, 30, tab.fcb11l, 3, 6, tab.cb1110l0, tab.cb1110l1, 20 }
    },
    512, 16, ff_metasound_lsp11, 1, 6, 4, 3, tab.shape11, 9, 36, 30, 7, 90
};

static const TwinVQModeTab mode_16_16 = {
    {
        { 8, bark_tab_s16_128,  10, tab.fcb16s, 1, 5, tab.cb1616s0, tab.cb1616s1, 16 },
        { 2, bark_tab_m16_512,  20, tab.fcb16m, 2, 5, tab.cb1616m0, tab.cb1616m1, 15 },
        { 1, bark_tab_l16_1024, 30, tab.fcb16l, 3, 6, tab.cb1616l0, tab.cb1616l1, 16 }
    },
    1024, 16, ff_metasound_lsp16, 1, 6, 4, 3, tab.shape16, 9, 56, 60, 7, 180
};

static const TwinVQModeTab mode_22_20 = {
    {
        { 8, bark_tab_s22_128,  10, tab.fcb22s_1, 1, 6, tab.cb2220s0, tab.cb2220s1, 18 },
        { 2, bark_tab_m22_512,  20, tab.fcb22m_1, 2, 6, tab.cb2220m0, tab.cb2220m1, 17 },
        { 1, bark_tab_l22_1024, 32, tab.fcb22l_1, 4, 6, tab.cb2220l0, tab.cb2220l1, 18 }
    },
    1024, 16, ff_metasound_lsp22, 1, 6, 4, 3, tab.shape22_1, 9, 56, 36, 7, 144
};

static const TwinVQModeTab mode_22_24 = {
    {
        { 8, bark_tab_s22_128,  10, tab.fcb22s_1, 1, 6, tab.cb2224s0, tab.cb2224s1, 15 },
        { 2, bark_tab_m22_512,  20, tab.fcb22m_1, 2, 6, tab.cb2224m0, tab.cb2224m1, 14 },
        { 1, bark_tab_l22_1024, 32, tab.fcb22l_1, 4, 6, tab.cb2224l0, tab.cb2224l1, 15 }
    },
    1024, 16, ff_metasound_lsp22, 1, 6, 4, 3, tab.shape22_1, 9, 56, 36, 7, 144
};

static const TwinVQModeTab mode_22_32 = {
    {
        { 4, bark_tab_s22_128, 10, tab.fcb22s_2, 1, 6, tab.cb2232s0, tab.cb2232s1, 11 },
        { 2, bark_tab_m22_256, 20, tab.fcb22m_2, 2, 6, tab.cb2232m0, tab.cb2232m1, 11 },
        { 1, bark_tab_l22_512, 32, tab.fcb22l_2, 4, 6, tab.cb2232l0, tab.cb2232l1, 12 }
    },
    512, 16, tab.lsp22_2, 1, 6, 4, 4, tab.shape22_2, 9, 56, 36, 7, 72
};

static const TwinVQModeTab mode_44_40 = {
    {
        { 16, bark_tab_s44_128,  10, tab.fcb44s, 1, 6, tab.cb4440s0, tab.cb4440s1, 18 },
        { 4,  bark_tab_m44_512,  20, tab.fcb44m, 2, 6, tab.cb4440m0, tab.cb4440m1, 17 },
        { 1,  bark_tab_l44_2048, 40, tab.fcb44l, 4, 6, tab.cb4440l0, tab.cb4440l1, 17 }
    },
    2048, 20, ff_metasound_lsp44, 1, 6, 4, 4, tab.shape44, 9, 84, 54, 7, 432
};

static const TwinVQModeTab mode_44_48 = {
    {
        { 16, bark_tab_s44_128,  10, tab.fcb44s, 1, 6, tab.cb4448s0, tab.cb4448s1, 15 },
        { 4,  bark_tab_m44_512,  20, tab.fcb44m, 2, 6, tab.cb4448m0, tab.cb4448m1, 14 },
        { 1,  bark_tab_l44_2048, 40, tab.fcb44l, 4, 6, tab.cb4448l0, tab.cb4448l1, 14 }
    },
    2048, 20, ff_metasound_lsp44, 1, 6, 4, 4, tab.shape44, 9, 84, 54, 7, 432
};

/**
 * Evaluate a * b / 400 rounded to the nearest integer. When, for example,
 * a * b == 200 and the nearest integer is ill-defined, use a table to emulate
 * the following broken float-based implementation used by the binary decoder:
 *
 * @code
 * static int very_broken_op(int a, int b)
 * {
 *    static float test; // Ugh, force gcc to do the division first...
 *
 *    test = a / 400.0;
 *    return b * test + 0.5;
 * }
 * @endcode
 *
 * @note if this function is replaced by just ROUNDED_DIV(a * b, 400.0), the
 * stddev between the original file (before encoding with Yamaha encoder) and
 * the decoded output increases, which leads one to believe that the encoder
 * expects exactly this broken calculation.
 */
static int very_broken_op(int a, int b)
{
    int x = a * b + 200;
    int size;
    const uint8_t *rtab;

    if (x % 400 || b % 5)
        return x / 400;

    x /= 400;

    size = tabs[b / 5].size;
    rtab = tabs[b / 5].tab;
    return x - rtab[size * av_log2(2 * (x - 1) / size) + (x - 1) % size];
}

/**
 * Sum to data a periodic peak of a given period, width and shape.
 *
 * @param period the period of the peak divided by 400.0
 */
static void add_peak(int period, int width, const float *shape,
                     float ppc_gain, float *speech, int len)
{
    int i, j;

    const float *shape_end = shape + len;
    int center;

    // First peak centered around zero
    for (i = 0; i < width / 2; i++)
        speech[i] += ppc_gain * *shape++;

    for (i = 1; i < ROUNDED_DIV(len, width); i++) {
        center = very_broken_op(period, i);
        for (j = -width / 2; j < (width + 1) / 2; j++)
            speech[j + center] += ppc_gain * *shape++;
    }

    // For the last block, be careful not to go beyond the end of the buffer
    center = very_broken_op(period, i);
    for (j = -width / 2; j < (width + 1) / 2 && shape < shape_end; j++)
        speech[j + center] += ppc_gain * *shape++;
}

static void decode_ppc(TwinVQContext *tctx, int period_coef, int g_coef,
                       const float *shape, float *speech)
{
    const TwinVQModeTab *mtab = tctx->mtab;
    int isampf = tctx->avctx->sample_rate /  1000;
    int ibps   = tctx->avctx->bit_rate    / (1000 * tctx->avctx->ch_layout.nb_channels);
    int min_period   = ROUNDED_DIV(40 * 2 * mtab->size, isampf);
    int max_period   = ROUNDED_DIV(40 * 2 * mtab->size * 6, isampf);
    int period_range = max_period - min_period;
    float pgain_step = 25000.0 / ((1 << mtab->pgain_bit) - 1);
    float ppc_gain   = 1.0 / 8192 *
                       twinvq_mulawinv(pgain_step * g_coef +
                                           pgain_step / 2,
                                       25000.0, TWINVQ_PGAIN_MU);

    // This is actually the period multiplied by 400. It is just linearly coded
    // between its maximum and minimum value.
    int period = min_period +
                 ROUNDED_DIV(period_coef * period_range,
                             (1 << mtab->ppc_period_bit) - 1);
    int width;

    if (isampf == 22 && ibps == 32) {
        // For some unknown reason, NTT decided to code this case differently...
        width = ROUNDED_DIV((period + 800) * mtab->peak_per2wid,
                            400 * mtab->size);
    } else
        width = period * mtab->peak_per2wid / (400 * mtab->size);

    add_peak(period, width, shape, ppc_gain, speech, mtab->ppc_shape_len);
}

static void dec_bark_env(TwinVQContext *tctx, const uint8_t *in, int use_hist,
                         int ch, float *out, float gain,
                         enum TwinVQFrameType ftype)
{
    const TwinVQModeTab *mtab = tctx->mtab;
    int i, j;
    float *hist     = tctx->bark_hist[ftype][ch];
    float val       = ((const float []) { 0.4, 0.35, 0.28 })[ftype];
    int bark_n_coef = mtab->fmode[ftype].bark_n_coef;
    int fw_cb_len   = mtab->fmode[ftype].bark_env_size / bark_n_coef;
    int idx         = 0;

    for (i = 0; i < fw_cb_len; i++)
        for (j = 0; j < bark_n_coef; j++, idx++) {
            float tmp2 = mtab->fmode[ftype].bark_cb[fw_cb_len * in[j] + i] *
                         (1.0 / 4096);
            float st   = use_hist ? (1.0 - val) * tmp2 + val * hist[idx] + 1.0
                                  : tmp2 + 1.0;

            hist[idx] = tmp2;
            if (st < -1.0)
                st = 1.0;

            twinvq_memset_float(out, st * gain, mtab->fmode[ftype].bark_tab[idx]);
            out += mtab->fmode[ftype].bark_tab[idx];
        }
}

static void read_cb_data(TwinVQContext *tctx, GetBitContext *gb,
                         uint8_t *dst, enum TwinVQFrameType ftype)
{
    int i;

    for (i = 0; i < tctx->n_div[ftype]; i++) {
        int bs_second_part = (i >= tctx->bits_main_spec_change[ftype]);

        *dst++ = get_bits(gb, tctx->bits_main_spec[0][ftype][bs_second_part]);
        *dst++ = get_bits(gb, tctx->bits_main_spec[1][ftype][bs_second_part]);
    }
}

static int twinvq_read_bitstream(AVCodecContext *avctx, TwinVQContext *tctx,
                                 const uint8_t *buf, int buf_size)
{
    TwinVQFrameData     *bits = &tctx->bits[0];
    const TwinVQModeTab *mtab = tctx->mtab;
    int channels              = tctx->avctx->ch_layout.nb_channels;
    int sub;
    GetBitContext gb;
    int i, j, k, ret;

    if ((ret = init_get_bits8(&gb, buf, buf_size)) < 0)
        return ret;
    skip_bits(&gb, get_bits(&gb, 8));

    bits->window_type = get_bits(&gb, TWINVQ_WINDOW_TYPE_BITS);

    if (bits->window_type > 8) {
        av_log(avctx, AV_LOG_ERROR, "Invalid window type, broken sample?\n");
        return AVERROR_INVALIDDATA;
    }

    bits->ftype = ff_twinvq_wtype_to_ftype_table[tctx->bits[0].window_type];

    sub = mtab->fmode[bits->ftype].sub;

    read_cb_data(tctx, &gb, bits->main_coeffs, bits->ftype);

    for (i = 0; i < channels; i++)
        for (j = 0; j < sub; j++)
            for (k = 0; k < mtab->fmode[bits->ftype].bark_n_coef; k++)
                bits->bark1[i][j][k] =
                    get_bits(&gb, mtab->fmode[bits->ftype].bark_n_bit);

    for (i = 0; i < channels; i++)
        for (j = 0; j < sub; j++)
            bits->bark_use_hist[i][j] = get_bits1(&gb);

    if (bits->ftype == TWINVQ_FT_LONG) {
        for (i = 0; i < channels; i++)
            bits->gain_bits[i] = get_bits(&gb, TWINVQ_GAIN_BITS);
    } else {
        for (i = 0; i < channels; i++) {
            bits->gain_bits[i] = get_bits(&gb, TWINVQ_GAIN_BITS);
            for (j = 0; j < sub; j++)
                bits->sub_gain_bits[i * sub + j] = get_bits(&gb,
                                                       TWINVQ_SUB_GAIN_BITS);
        }
    }

    for (i = 0; i < channels; i++) {
        bits->lpc_hist_idx[i] = get_bits(&gb, mtab->lsp_bit0);
        bits->lpc_idx1[i]     = get_bits(&gb, mtab->lsp_bit1);

        for (j = 0; j < mtab->lsp_split; j++)
            bits->lpc_idx2[i][j] = get_bits(&gb, mtab->lsp_bit2);
    }

    if (bits->ftype == TWINVQ_FT_LONG) {
        read_cb_data(tctx, &gb, bits->ppc_coeffs, 3);
        for (i = 0; i < channels; i++) {
            bits->p_coef[i] = get_bits(&gb, mtab->ppc_period_bit);
            bits->g_coef[i] = get_bits(&gb, mtab->pgain_bit);
        }
    }

    return (get_bits_count(&gb) + 7) / 8;
}

static av_cold int twinvq_decode_init(AVCodecContext *avctx)
{
    int isampf, ibps, channels;
    TwinVQContext *tctx = avctx->priv_data;

    if (!avctx->extradata || avctx->extradata_size < 12) {
        av_log(avctx, AV_LOG_ERROR, "Missing or incomplete extradata\n");
        return AVERROR_INVALIDDATA;
    }
    channels        = AV_RB32(avctx->extradata)     + 1;
    avctx->bit_rate = AV_RB32(avctx->extradata + 4) * 1000;
    isampf          = AV_RB32(avctx->extradata + 8);

    if (isampf < 8 || isampf > 44) {
        av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate\n");
        return AVERROR_INVALIDDATA;
    }
    switch (isampf) {
    case 44:
        avctx->sample_rate = 44100;
        break;
    case 22:
        avctx->sample_rate = 22050;
        break;
    case 11:
        avctx->sample_rate = 11025;
        break;
    default:
        avctx->sample_rate = isampf * 1000;
        break;
    }

    if (channels <= 0 || channels > TWINVQ_CHANNELS_MAX) {
        av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %i\n",
               channels);
        return -1;
    }
    av_channel_layout_uninit(&avctx->ch_layout);
    av_channel_layout_default(&avctx->ch_layout, channels);

    ibps = avctx->bit_rate / (1000 * channels);
    if (ibps < 8 || ibps > 48) {
        av_log(avctx, AV_LOG_ERROR, "Bad bitrate per channel value %d\n", ibps);
        return AVERROR_INVALIDDATA;
    }

    switch ((isampf << 8) + ibps) {
    case (8 << 8) + 8:
        tctx->mtab = &mode_08_08;
        break;
    case (11 << 8) + 8:
        tctx->mtab = &mode_11_08;
        break;
    case (11 << 8) + 10:
        tctx->mtab = &mode_11_10;
        break;
    case (16 << 8) + 16:
        tctx->mtab = &mode_16_16;
        break;
    case (22 << 8) + 20:
        tctx->mtab = &mode_22_20;
        break;
    case (22 << 8) + 24:
        tctx->mtab = &mode_22_24;
        break;
    case (22 << 8) + 32:
        tctx->mtab = &mode_22_32;
        break;
    case (44 << 8) + 40:
        tctx->mtab = &mode_44_40;
        break;
    case (44 << 8) + 48:
        tctx->mtab = &mode_44_48;
        break;
    default:
        av_log(avctx, AV_LOG_ERROR,
               "This version does not support %d kHz - %d kbit/s/ch mode.\n",
               isampf, isampf);
        return -1;
    }

    tctx->codec          = TWINVQ_CODEC_VQF;
    tctx->read_bitstream = twinvq_read_bitstream;
    tctx->dec_bark_env   = dec_bark_env;
    tctx->decode_ppc     = decode_ppc;
    tctx->frame_size     = avctx->bit_rate * tctx->mtab->size
                                           / avctx->sample_rate + 8;
    tctx->is_6kbps       = 0;
    if (avctx->block_align && avctx->block_align * 8LL / tctx->frame_size > 1) {
        av_log(avctx, AV_LOG_ERROR,
               "VQF TwinVQ should have only one frame per packet\n");
        return AVERROR_INVALIDDATA;
    }

    return ff_twinvq_decode_init(avctx);
}

const FFCodec ff_twinvq_decoder = {
    .p.name         = "twinvq",
    CODEC_LONG_NAME("VQF TwinVQ"),
    .p.type         = AVMEDIA_TYPE_AUDIO,
    .p.id           = AV_CODEC_ID_TWINVQ,
    .priv_data_size = sizeof(TwinVQContext),
    .init           = twinvq_decode_init,
    .close          = ff_twinvq_decode_close,
    FF_CODEC_DECODE_CB(ff_twinvq_decode_frame),
    .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_CHANNEL_CONF,
    .p.sample_fmts  = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
                                                      AV_SAMPLE_FMT_NONE },
    .caps_internal  = FF_CODEC_CAP_INIT_CLEANUP,
};