FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
binkaudio.c
Go to the documentation of this file.
1 /*
2  * Bink Audio decoder
3  * Copyright (c) 2007-2011 Peter Ross (pross@xvid.org)
4  * Copyright (c) 2009 Daniel Verkamp (daniel@drv.nu)
5  *
6  * This file is part of FFmpeg.
7  *
8  * FFmpeg is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * FFmpeg is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with FFmpeg; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
23 /**
24  * @file
25  * Bink Audio decoder
26  *
27  * Technical details here:
28  * http://wiki.multimedia.cx/index.php?title=Bink_Audio
29  */
30 
32 #include "avcodec.h"
33 #define BITSTREAM_READER_LE
34 #include "get_bits.h"
35 #include "dct.h"
36 #include "rdft.h"
37 #include "fmtconvert.h"
38 #include "internal.h"
39 #include "libavutil/intfloat.h"
40 
41 extern const uint16_t ff_wma_critical_freqs[25];
42 
43 static float quant_table[96];
44 
45 #define MAX_CHANNELS 2
46 #define BINK_BLOCK_MAX_SIZE (MAX_CHANNELS << 11)
47 
48 typedef struct {
50  int version_b; ///< Bink version 'b'
51  int first;
52  int channels;
53  int frame_len; ///< transform size (samples)
54  int overlap_len; ///< overlap size (samples)
56  int num_bands;
57  unsigned int *bands;
58  float root;
60  float previous[MAX_CHANNELS][BINK_BLOCK_MAX_SIZE / 16]; ///< coeffs from previous audio block
62  union {
65  } trans;
67 
68 
70 {
71  BinkAudioContext *s = avctx->priv_data;
72  int sample_rate = avctx->sample_rate;
73  int sample_rate_half;
74  int i;
75  int frame_len_bits;
76 
77  /* determine frame length */
78  if (avctx->sample_rate < 22050) {
79  frame_len_bits = 9;
80  } else if (avctx->sample_rate < 44100) {
81  frame_len_bits = 10;
82  } else {
83  frame_len_bits = 11;
84  }
85 
86  if (avctx->channels < 1 || avctx->channels > MAX_CHANNELS) {
87  av_log(avctx, AV_LOG_ERROR, "invalid number of channels: %d\n", avctx->channels);
88  return AVERROR_INVALIDDATA;
89  }
90  avctx->channel_layout = avctx->channels == 1 ? AV_CH_LAYOUT_MONO :
92 
93  s->version_b = avctx->extradata_size >= 4 && avctx->extradata[3] == 'b';
94 
95  if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT) {
96  // audio is already interleaved for the RDFT format variant
98  sample_rate *= avctx->channels;
99  s->channels = 1;
100  if (!s->version_b)
101  frame_len_bits += av_log2(avctx->channels);
102  } else {
103  s->channels = avctx->channels;
105  }
106 
107  s->frame_len = 1 << frame_len_bits;
108  s->overlap_len = s->frame_len / 16;
109  s->block_size = (s->frame_len - s->overlap_len) * s->channels;
110  sample_rate_half = (sample_rate + 1) / 2;
111  if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
112  s->root = 2.0 / (sqrt(s->frame_len) * 32768.0);
113  else
114  s->root = s->frame_len / (sqrt(s->frame_len) * 32768.0);
115  for (i = 0; i < 96; i++) {
116  /* constant is result of 0.066399999/log10(M_E) */
117  quant_table[i] = expf(i * 0.15289164787221953823f) * s->root;
118  }
119 
120  /* calculate number of bands */
121  for (s->num_bands = 1; s->num_bands < 25; s->num_bands++)
122  if (sample_rate_half <= ff_wma_critical_freqs[s->num_bands - 1])
123  break;
124 
125  s->bands = av_malloc((s->num_bands + 1) * sizeof(*s->bands));
126  if (!s->bands)
127  return AVERROR(ENOMEM);
128 
129  /* populate bands data */
130  s->bands[0] = 2;
131  for (i = 1; i < s->num_bands; i++)
132  s->bands[i] = (ff_wma_critical_freqs[i - 1] * s->frame_len / sample_rate_half) & ~1;
133  s->bands[s->num_bands] = s->frame_len;
134 
135  s->first = 1;
136 
137  if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
138  ff_rdft_init(&s->trans.rdft, frame_len_bits, DFT_C2R);
139  else if (CONFIG_BINKAUDIO_DCT_DECODER)
140  ff_dct_init(&s->trans.dct, frame_len_bits, DCT_III);
141  else
142  return -1;
143 
144  return 0;
145 }
146 
147 static float get_float(GetBitContext *gb)
148 {
149  int power = get_bits(gb, 5);
150  float f = ldexpf(get_bits_long(gb, 23), power - 23);
151  if (get_bits1(gb))
152  f = -f;
153  return f;
154 }
155 
156 static const uint8_t rle_length_tab[16] = {
157  2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 32, 64
158 };
159 
160 /**
161  * Decode Bink Audio block
162  * @param[out] out Output buffer (must contain s->block_size elements)
163  * @return 0 on success, negative error code on failure
164  */
165 static int decode_block(BinkAudioContext *s, float **out, int use_dct)
166 {
167  int ch, i, j, k;
168  float q, quant[25];
169  int width, coeff;
170  GetBitContext *gb = &s->gb;
171 
172  if (use_dct)
173  skip_bits(gb, 2);
174 
175  for (ch = 0; ch < s->channels; ch++) {
176  FFTSample *coeffs = out[ch];
177 
178  if (s->version_b) {
179  if (get_bits_left(gb) < 64)
180  return AVERROR_INVALIDDATA;
181  coeffs[0] = av_int2float(get_bits_long(gb, 32)) * s->root;
182  coeffs[1] = av_int2float(get_bits_long(gb, 32)) * s->root;
183  } else {
184  if (get_bits_left(gb) < 58)
185  return AVERROR_INVALIDDATA;
186  coeffs[0] = get_float(gb) * s->root;
187  coeffs[1] = get_float(gb) * s->root;
188  }
189 
190  if (get_bits_left(gb) < s->num_bands * 8)
191  return AVERROR_INVALIDDATA;
192  for (i = 0; i < s->num_bands; i++) {
193  int value = get_bits(gb, 8);
194  quant[i] = quant_table[FFMIN(value, 95)];
195  }
196 
197  k = 0;
198  q = quant[0];
199 
200  // parse coefficients
201  i = 2;
202  while (i < s->frame_len) {
203  if (s->version_b) {
204  j = i + 16;
205  } else {
206  int v = get_bits1(gb);
207  if (v) {
208  v = get_bits(gb, 4);
209  j = i + rle_length_tab[v] * 8;
210  } else {
211  j = i + 8;
212  }
213  }
214 
215  j = FFMIN(j, s->frame_len);
216 
217  width = get_bits(gb, 4);
218  if (width == 0) {
219  memset(coeffs + i, 0, (j - i) * sizeof(*coeffs));
220  i = j;
221  while (s->bands[k] < i)
222  q = quant[k++];
223  } else {
224  while (i < j) {
225  if (s->bands[k] == i)
226  q = quant[k++];
227  coeff = get_bits(gb, width);
228  if (coeff) {
229  int v;
230  v = get_bits1(gb);
231  if (v)
232  coeffs[i] = -q * coeff;
233  else
234  coeffs[i] = q * coeff;
235  } else {
236  coeffs[i] = 0.0f;
237  }
238  i++;
239  }
240  }
241  }
242 
243  if (CONFIG_BINKAUDIO_DCT_DECODER && use_dct) {
244  coeffs[0] /= 0.5;
245  s->trans.dct.dct_calc(&s->trans.dct, coeffs);
246  }
247  else if (CONFIG_BINKAUDIO_RDFT_DECODER)
248  s->trans.rdft.rdft_calc(&s->trans.rdft, coeffs);
249  }
250 
251  for (ch = 0; ch < s->channels; ch++) {
252  int j;
253  int count = s->overlap_len * s->channels;
254  if (!s->first) {
255  j = ch;
256  for (i = 0; i < s->overlap_len; i++, j += s->channels)
257  out[ch][i] = (s->previous[ch][i] * (count - j) +
258  out[ch][i] * j) / count;
259  }
260  memcpy(s->previous[ch], &out[ch][s->frame_len - s->overlap_len],
261  s->overlap_len * sizeof(*s->previous[ch]));
262  }
263 
264  s->first = 0;
265 
266  return 0;
267 }
268 
270 {
271  BinkAudioContext * s = avctx->priv_data;
272  av_freep(&s->bands);
273  av_freep(&s->packet_buffer);
274  if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
275  ff_rdft_end(&s->trans.rdft);
276  else if (CONFIG_BINKAUDIO_DCT_DECODER)
277  ff_dct_end(&s->trans.dct);
278 
279  return 0;
280 }
281 
283 {
284  int n = (-get_bits_count(s)) & 31;
285  if (n) skip_bits(s, n);
286 }
287 
288 static int decode_frame(AVCodecContext *avctx, void *data,
289  int *got_frame_ptr, AVPacket *avpkt)
290 {
291  BinkAudioContext *s = avctx->priv_data;
292  AVFrame *frame = data;
293  GetBitContext *gb = &s->gb;
294  int ret, consumed = 0;
295 
296  if (!get_bits_left(gb)) {
297  uint8_t *buf;
298  /* handle end-of-stream */
299  if (!avpkt->size) {
300  *got_frame_ptr = 0;
301  return 0;
302  }
303  if (avpkt->size < 4) {
304  av_log(avctx, AV_LOG_ERROR, "Packet is too small\n");
305  return AVERROR_INVALIDDATA;
306  }
308  if (!buf)
309  return AVERROR(ENOMEM);
310  s->packet_buffer = buf;
311  memcpy(s->packet_buffer, avpkt->data, avpkt->size);
312  init_get_bits(gb, s->packet_buffer, avpkt->size * 8);
313  consumed = avpkt->size;
314 
315  /* skip reported size */
316  skip_bits_long(gb, 32);
317  }
318 
319  /* get output buffer */
320  frame->nb_samples = s->frame_len;
321  if ((ret = ff_get_buffer(avctx, frame)) < 0) {
322  av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
323  return ret;
324  }
325 
326  if (decode_block(s, (float **)frame->extended_data,
327  avctx->codec->id == AV_CODEC_ID_BINKAUDIO_DCT)) {
328  av_log(avctx, AV_LOG_ERROR, "Incomplete packet\n");
329  return AVERROR_INVALIDDATA;
330  }
331  get_bits_align32(gb);
332 
333  frame->nb_samples = s->block_size / avctx->channels;
334  *got_frame_ptr = 1;
335 
336  return consumed;
337 }
338 
340  .name = "binkaudio_rdft",
341  .type = AVMEDIA_TYPE_AUDIO,
343  .priv_data_size = sizeof(BinkAudioContext),
344  .init = decode_init,
345  .close = decode_end,
346  .decode = decode_frame,
347  .capabilities = CODEC_CAP_DELAY | CODEC_CAP_DR1,
348  .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (RDFT)")
349 };
350 
352  .name = "binkaudio_dct",
353  .type = AVMEDIA_TYPE_AUDIO,
355  .priv_data_size = sizeof(BinkAudioContext),
356  .init = decode_init,
357  .close = decode_end,
358  .decode = decode_frame,
359  .capabilities = CODEC_CAP_DELAY | CODEC_CAP_DR1,
360  .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (DCT)")
361 };