FFmpeg coverage

Line	Branch	Exec	Source
1			/*
2			* Voxware MetaSound decoder
3			* Copyright (c) 2013 Konstantin Shishkov
4			* based on TwinVQ decoder
5			* Copyright (c) 2009 Vitor Sessak
6			*
7			* This file is part of FFmpeg.
8			*
9			* FFmpeg is free software; you can redistribute it and/or
10			* modify it under the terms of the GNU Lesser General Public
11			* License as published by the Free Software Foundation; either
12			* version 2.1 of the License, or (at your option) any later version.
13			*
14			* FFmpeg is distributed in the hope that it will be useful,
15			* but WITHOUT ANY WARRANTY; without even the implied warranty of
16			* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17			* Lesser General Public License for more details.
18			*
19			* You should have received a copy of the GNU Lesser General Public
20			* License along with FFmpeg; if not, write to the Free Software
21			* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22			*/
23
24			#include <inttypes.h>
25			#include <math.h>
26			#include <stdint.h>
27
28			#include "libavutil/channel_layout.h"
29
30			#define BITSTREAM_READER_LE
31			#include "avcodec.h"
32			#include "codec_internal.h"
33			#include "get_bits.h"
34
35			#include "twinvq.h"
36			#include "metasound_data.h"
37
38		✗	static void add_peak(float period, int width, const float *shape,
39			float ppc_gain, float *speech, int len)
40			{
41			int i, j, center;
42		✗	const float *shape_end = shape + len;
43
44			// First peak centered around zero
45		✗	for (i = 0; i < width / 2; i++)
46		✗	speech[i] += ppc_gain * *shape++;
47
48		✗	for (i = 1; i < ROUNDED_DIV(len, width); i++) {
49		✗	center = (int)(i * period + 0.5);
50		✗	for (j = -width / 2; j < (width + 1) / 2; j++)
51		✗	speech[j + center] += ppc_gain * *shape++;
52			}
53
54			// For the last block, be careful not to go beyond the end of the buffer
55		✗	center = (int)(i * period + 0.5);
56		✗	for (j = -width / 2; j < (width + 1) / 2 && shape < shape_end; j++)
57		✗	speech[j + center] += ppc_gain * *shape++;
58		✗	}
59
60		✗	static void decode_ppc(TwinVQContext *tctx, int period_coef, int g_coef,
61			const float *shape, float *speech)
62			{
63		✗	const TwinVQModeTab *mtab = tctx->mtab;
64		✗	int channels = tctx->avctx->ch_layout.nb_channels;
65		✗	int isampf = tctx->avctx->sample_rate / 1000;
66		✗	int ibps = tctx->avctx->bit_rate / (1000 * channels);
67			int width;
68
69		✗	float ratio = (float)mtab->size / isampf;
70			float min_period, max_period, period_range, period;
71			float some_mult;
72
73			float pgain_base, pgain_step, ppc_gain;
74
75		✗	if (channels == 1) {
76		✗	min_period = log2(ratio * 0.2);
77		✗	max_period = min_period + log2(6);
78			} else {
79		✗	min_period = (int)(ratio * 0.2 * 400 + 0.5) / 400.0;
80		✗	max_period = (int)(ratio * 0.2 * 400 * 6 + 0.5) / 400.0;
81			}
82		✗	period_range = max_period - min_period;
83		✗	period = min_period + period_coef * period_range /
84		✗	((1 << mtab->ppc_period_bit) - 1);
85		✗	if (channels == 1)
86		✗	period = powf(2.0, period);
87			else
88		✗	period = (int)(period * 400 + 0.5) / 400.0;
89
90		✗	switch (isampf) {
91		✗	case 8: some_mult = 2.0; break;
92		✗	case 11: some_mult = 3.0; break;
93		✗	case 16: some_mult = 3.0; break;
94		✗	case 22: some_mult = ibps == 32 ? 2.0 : 4.0; break;
95		✗	case 44: some_mult = 8.0; break;
96		✗	default: some_mult = 4.0;
97			}
98
99		✗	width = (int)(some_mult / (mtab->size / period) * mtab->ppc_shape_len);
100		✗	if (isampf == 22 && ibps == 32)
101		✗	width = (int)((2.0 / period + 1) * width + 0.5);
102
103		✗	pgain_base = channels == 2 ? 25000.0 : 20000.0;
104		✗	pgain_step = pgain_base / ((1 << mtab->pgain_bit) - 1);
105		✗	ppc_gain = 1.0 / 8192 *
106		✗	twinvq_mulawinv(pgain_step * g_coef + pgain_step / 2,
107			pgain_base, TWINVQ_PGAIN_MU);
108
109		✗	add_peak(period, width, shape, ppc_gain, speech, mtab->ppc_shape_len);
110		✗	}
111
112		✗	static void dec_bark_env(TwinVQContext *tctx, const uint8_t *in, int use_hist,
113			int ch, float *out, float gain,
114			enum TwinVQFrameType ftype)
115			{
116		✗	const TwinVQModeTab *mtab = tctx->mtab;
117			int i, j;
118		✗	float *hist = tctx->bark_hist[ftype][ch];
119		✗	float val = ((const float []) { 0.4, 0.35, 0.28 })[ftype];
120		✗	int bark_n_coef = mtab->fmode[ftype].bark_n_coef;
121		✗	int fw_cb_len = mtab->fmode[ftype].bark_env_size / bark_n_coef;
122		✗	int idx = 0;
123		✗	int channels = tctx->avctx->ch_layout.nb_channels;
124
125		✗	if (channels == 1)
126		✗	val = 0.5;
127		✗	for (i = 0; i < fw_cb_len; i++)
128		✗	for (j = 0; j < bark_n_coef; j++, idx++) {
129		✗	float tmp2 = mtab->fmode[ftype].bark_cb[fw_cb_len * in[j] + i] *
130			(1.0 / 2048);
131			float st;
132
133		✗	if (channels == 1)
134		✗	st = use_hist ?
135		✗	tmp2 + val * hist[idx] + 1.0 : tmp2 + 1.0;
136			else
137		✗	st = use_hist ? (1.0 - val) * tmp2 + val * hist[idx] + 1.0
138		✗	: tmp2 + 1.0;
139
140		✗	hist[idx] = tmp2;
141		✗	if (st < 0.1)
142		✗	st = 0.1;
143
144		✗	twinvq_memset_float(out, st * gain,
145		✗	mtab->fmode[ftype].bark_tab[idx]);
146		✗	out += mtab->fmode[ftype].bark_tab[idx];
147			}
148		✗	}
149
150		✗	static void read_cb_data(TwinVQContext *tctx, GetBitContext *gb,
151			uint8_t *dst, enum TwinVQFrameType ftype)
152			{
153			int i;
154
155		✗	for (i = 0; i < tctx->n_div[ftype]; i++) {
156		✗	int bs_second_part = (i >= tctx->bits_main_spec_change[ftype]);
157
158		✗	*dst++ = get_bits(gb, tctx->bits_main_spec[0][ftype][bs_second_part]);
159		✗	*dst++ = get_bits(gb, tctx->bits_main_spec[1][ftype][bs_second_part]);
160			}
161		✗	}
162
163		✗	static int metasound_read_bitstream(AVCodecContext *avctx, TwinVQContext *tctx,
164			const uint8_t *buf, int buf_size)
165			{
166			TwinVQFrameData *bits;
167		✗	const TwinVQModeTab *mtab = tctx->mtab;
168		✗	int channels = tctx->avctx->ch_layout.nb_channels;
169			int sub;
170			GetBitContext gb;
171			int i, j, k, ret;
172
173		✗	if ((ret = init_get_bits8(&gb, buf, buf_size)) < 0)
174		✗	return ret;
175
176		✗	for (tctx->cur_frame = 0; tctx->cur_frame < tctx->frames_per_packet;
177		✗	tctx->cur_frame++) {
178		✗	bits = tctx->bits + tctx->cur_frame;
179
180		✗	bits->window_type = get_bits(&gb, TWINVQ_WINDOW_TYPE_BITS);
181
182		✗	if (bits->window_type > 8) {
183		✗	av_log(avctx, AV_LOG_ERROR, "Invalid window type, broken sample?\n");
184		✗	return AVERROR_INVALIDDATA;
185			}
186
187		✗	bits->ftype = ff_twinvq_wtype_to_ftype_table[tctx->bits[tctx->cur_frame].window_type];
188
189		✗	sub = mtab->fmode[bits->ftype].sub;
190
191		✗	if (bits->ftype != TWINVQ_FT_SHORT && !tctx->is_6kbps)
192		✗	get_bits(&gb, 2);
193
194		✗	read_cb_data(tctx, &gb, bits->main_coeffs, bits->ftype);
195
196		✗	for (i = 0; i < channels; i++)
197		✗	for (j = 0; j < sub; j++)
198		✗	for (k = 0; k < mtab->fmode[bits->ftype].bark_n_coef; k++)
199		✗	bits->bark1[i][j][k] =
200		✗	get_bits(&gb, mtab->fmode[bits->ftype].bark_n_bit);
201
202		✗	for (i = 0; i < channels; i++)
203		✗	for (j = 0; j < sub; j++)
204		✗	bits->bark_use_hist[i][j] = get_bits1(&gb);
205
206		✗	if (bits->ftype == TWINVQ_FT_LONG) {
207		✗	for (i = 0; i < channels; i++)
208		✗	bits->gain_bits[i] = get_bits(&gb, TWINVQ_GAIN_BITS);
209			} else {
210		✗	for (i = 0; i < channels; i++) {
211		✗	bits->gain_bits[i] = get_bits(&gb, TWINVQ_GAIN_BITS);
212		✗	for (j = 0; j < sub; j++)
213		✗	bits->sub_gain_bits[i * sub + j] =
214		✗	get_bits(&gb, TWINVQ_SUB_GAIN_BITS);
215			}
216			}
217
218		✗	for (i = 0; i < channels; i++) {
219		✗	bits->lpc_hist_idx[i] = get_bits(&gb, mtab->lsp_bit0);
220		✗	bits->lpc_idx1[i] = get_bits(&gb, mtab->lsp_bit1);
221
222		✗	for (j = 0; j < mtab->lsp_split; j++)
223		✗	bits->lpc_idx2[i][j] = get_bits(&gb, mtab->lsp_bit2);
224			}
225
226		✗	if (bits->ftype == TWINVQ_FT_LONG) {
227		✗	read_cb_data(tctx, &gb, bits->ppc_coeffs, 3);
228		✗	for (i = 0; i < channels; i++) {
229		✗	bits->p_coef[i] = get_bits(&gb, mtab->ppc_period_bit);
230		✗	bits->g_coef[i] = get_bits(&gb, mtab->pgain_bit);
231			}
232			}
233
234			// subframes are aligned to nibbles
235		✗	if (get_bits_count(&gb) & 3)
236		✗	skip_bits(&gb, 4 - (get_bits_count(&gb) & 3));
237			}
238
239		✗	return (get_bits_count(&gb) + 7) / 8;
240			}
241
242			typedef struct MetasoundProps {
243			uint32_t tag;
244			int bit_rate;
245			int channels;
246			int sample_rate;
247			} MetasoundProps;
248
249			static const MetasoundProps codec_props[] = {
250			{ MKTAG('V','X','0','3'), 6, 1, 8000 },
251			{ MKTAG('V','X','0','4'), 12, 2, 8000 },
252
253			{ MKTAG('V','O','X','i'), 8, 1, 8000 },
254			{ MKTAG('V','O','X','j'), 10, 1, 11025 },
255			{ MKTAG('V','O','X','k'), 16, 1, 16000 },
256			{ MKTAG('V','O','X','L'), 24, 1, 22050 },
257			{ MKTAG('V','O','X','q'), 32, 1, 44100 },
258			{ MKTAG('V','O','X','r'), 40, 1, 44100 },
259			{ MKTAG('V','O','X','s'), 48, 1, 44100 },
260			{ MKTAG('V','O','X','t'), 16, 2, 8000 },
261			{ MKTAG('V','O','X','u'), 20, 2, 11025 },
262			{ MKTAG('V','O','X','v'), 32, 2, 16000 },
263			{ MKTAG('V','O','X','w'), 48, 2, 22050 },
264			{ MKTAG('V','O','X','x'), 64, 2, 44100 },
265			{ MKTAG('V','O','X','y'), 80, 2, 44100 },
266			{ MKTAG('V','O','X','z'), 96, 2, 44100 },
267
268			{ 0, 0, 0, 0 }
269			};
270
271		✗	static av_cold int metasound_decode_init(AVCodecContext *avctx)
272			{
273			int isampf, ibps;
274		✗	TwinVQContext *tctx = avctx->priv_data;
275			uint32_t tag;
276		✗	const MetasoundProps *props = codec_props;
277			int channels;
278
279		✗	if (!avctx->extradata || avctx->extradata_size < 16) {
280		✗	av_log(avctx, AV_LOG_ERROR, "Missing or incomplete extradata\n");
281		✗	return AVERROR_INVALIDDATA;
282			}
283
284		✗	tag = AV_RL32(avctx->extradata + 12);
285
286			for (;;) {
287		✗	if (!props->tag) {
288		✗	av_log(avctx, AV_LOG_ERROR, "Could not find tag %08"PRIX32"\n", tag);
289		✗	return AVERROR_INVALIDDATA;
290			}
291		✗	if (props->tag == tag) {
292		✗	avctx->sample_rate = props->sample_rate;
293		✗	channels = props->channels;
294		✗	avctx->bit_rate = props->bit_rate * 1000;
295		✗	isampf = avctx->sample_rate / 1000;
296		✗	break;
297			}
298		✗	props++;
299			}
300
301		✗	av_channel_layout_uninit(&avctx->ch_layout);
302		✗	av_channel_layout_default(&avctx->ch_layout, channels);
303
304		✗	ibps = avctx->bit_rate / (1000 * channels);
305
306		✗	switch ((channels << 16) + (isampf << 8) + ibps) {
307		✗	case (1 << 16) + ( 8 << 8) + 6:
308		✗	tctx->mtab = &metasound_mode0806;
309		✗	break;
310		✗	case (2 << 16) + ( 8 << 8) + 6:
311		✗	tctx->mtab = &metasound_mode0806s;
312		✗	break;
313		✗	case (1 << 16) + ( 8 << 8) + 8:
314		✗	tctx->mtab = &metasound_mode0808;
315		✗	break;
316		✗	case (2 << 16) + ( 8 << 8) + 8:
317		✗	tctx->mtab = &metasound_mode0808s;
318		✗	break;
319		✗	case (1 << 16) + (11 << 8) + 10:
320		✗	tctx->mtab = &metasound_mode1110;
321		✗	break;
322		✗	case (2 << 16) + (11 << 8) + 10:
323		✗	tctx->mtab = &metasound_mode1110s;
324		✗	break;
325		✗	case (1 << 16) + (16 << 8) + 16:
326		✗	tctx->mtab = &metasound_mode1616;
327		✗	break;
328		✗	case (2 << 16) + (16 << 8) + 16:
329		✗	tctx->mtab = &metasound_mode1616s;
330		✗	break;
331		✗	case (1 << 16) + (22 << 8) + 24:
332		✗	tctx->mtab = &metasound_mode2224;
333		✗	break;
334		✗	case (2 << 16) + (22 << 8) + 24:
335		✗	tctx->mtab = &metasound_mode2224s;
336		✗	break;
337		✗	case (1 << 16) + (44 << 8) + 32:
338			case (2 << 16) + (44 << 8) + 32:
339		✗	tctx->mtab = &metasound_mode4432;
340		✗	break;
341		✗	case (1 << 16) + (44 << 8) + 40:
342			case (2 << 16) + (44 << 8) + 40:
343		✗	tctx->mtab = &metasound_mode4440;
344		✗	break;
345		✗	case (1 << 16) + (44 << 8) + 48:
346			case (2 << 16) + (44 << 8) + 48:
347		✗	tctx->mtab = &metasound_mode4448;
348		✗	break;
349		✗	default:
350		✗	av_log(avctx, AV_LOG_ERROR,
351			"This version does not support %d kHz - %d kbit/s/ch mode.\n",
352			isampf, ibps);
353		✗	return AVERROR(ENOSYS);
354			}
355
356		✗	tctx->codec = TWINVQ_CODEC_METASOUND;
357		✗	tctx->read_bitstream = metasound_read_bitstream;
358		✗	tctx->dec_bark_env = dec_bark_env;
359		✗	tctx->decode_ppc = decode_ppc;
360		✗	tctx->frame_size = avctx->bit_rate * tctx->mtab->size
361		✗	/ avctx->sample_rate;
362		✗	tctx->is_6kbps = ibps == 6;
363
364		✗	return ff_twinvq_decode_init(avctx);
365			}
366
367			const FFCodec ff_metasound_decoder = {
368			.p.name = "metasound",
369			CODEC_LONG_NAME("Voxware MetaSound"),
370			.p.type = AVMEDIA_TYPE_AUDIO,
371			.p.id = AV_CODEC_ID_METASOUND,
372			.priv_data_size = sizeof(TwinVQContext),
373			.init = metasound_decode_init,
374			.close = ff_twinvq_decode_close,
375			FF_CODEC_DECODE_CB(ff_twinvq_decode_frame),
376			.p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_CHANNEL_CONF,
377			CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_FLTP),
378			.caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
379			};
380