Mercurial > libavcodec.hg
comparison nellymoserdec.c @ 5823:4beef9d0e663 libavcodec
Nellymoser ASAO decoder
author | banan |
---|---|
date | Mon, 15 Oct 2007 16:26:02 +0000 |
parents | |
children | fd8b56703750 |
comparison
equal
deleted
inserted
replaced
5822:b49be74d387c | 5823:4beef9d0e663 |
---|---|
1 /* | |
2 * NellyMoser audio decoder | |
3 * Copyright (c) 2007 a840bda5870ba11f19698ff6eb9581dfb0f95fa5, | |
4 * 539459aeb7d425140b62a3ec7dbf6dc8e408a306, and | |
5 * 520e17cd55896441042b14df2566a6eb610ed444 | |
6 * Copyright (c) 2007 Loic Minier <lool at dooz.org> | |
7 * Benjamin Larsson | |
8 * | |
9 * Permission is hereby granted, free of charge, to any person obtaining a | |
10 * copy of this software and associated documentation files (the "Software"), | |
11 * to deal in the Software without restriction, including without limitation | |
12 * the rights to use, copy, modify, merge, publish, distribute, sublicense, | |
13 * and/or sell copies of the Software, and to permit persons to whom the | |
14 * Software is furnished to do so, subject to the following conditions: | |
15 * | |
16 * The above copyright notice and this permission notice shall be included in | |
17 * all copies or substantial portions of the Software. | |
18 * | |
19 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
20 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
21 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
22 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
23 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING | |
24 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER | |
25 * DEALINGS IN THE SOFTWARE. | |
26 */ | |
27 | |
28 /** | |
29 * @file nellymoserdec.c | |
30 * The 3 alphanumeric copyright notices are md5summed they are from the original | |
31 * implementors. The original code is available from http://code.google.com/p/nelly2pcm/ | |
32 */ | |
33 #include "avcodec.h" | |
34 #include "random.h" | |
35 #include "dsputil.h" | |
36 | |
37 #define ALT_BITSTREAM_READER_LE | |
38 #include "bitstream.h" | |
39 | |
40 #define NELLY_BANDS 23 | |
41 #define NELLY_BLOCK_LEN 64 | |
42 #define NELLY_HEADER_BITS 116 | |
43 #define NELLY_DETAIL_BITS 198 | |
44 #define NELLY_BUF_LEN 128 | |
45 #define NELLY_FILL_LEN 124 | |
46 #define NELLY_BIT_CAP 6 | |
47 #define NELLY_BASE_OFF 4228 | |
48 #define NELLY_BASE_SHIFT 19 | |
49 #define NELLY_SAMPLES 256 | |
50 | |
51 static const float dequantization_table[127] = { | |
52 0.0000000000,-0.8472560048, 0.7224709988, -1.5247479677, -0.4531480074, 0.3753609955, 1.4717899561, | |
53 -1.9822579622, -1.1929379702, -0.5829370022, -0.0693780035, 0.3909569979,0.9069200158, 1.4862740040, | |
54 2.2215409279, -2.3887870312, -1.8067539930, -1.4105420113, -1.0773609877, -0.7995010018,-0.5558109879, | |
55 -0.3334020078, -0.1324490011, 0.0568020009, 0.2548770010, 0.4773550034, 0.7386850119, 1.0443060398, | |
56 1.3954459429, 1.8098750114, 2.3918759823,-2.3893830776, -1.9884680510, -1.7514040470, -1.5643119812, | |
57 -1.3922129869,-1.2164649963, -1.0469499826, -0.8905100226, -0.7645580173, -0.6454579830, -0.5259280205, | |
58 -0.4059549868, -0.3029719889, -0.2096900046, -0.1239869967, -0.0479229987, 0.0257730000, 0.1001340002, | |
59 0.1737180054, 0.2585540116, 0.3522900045, 0.4569880068, 0.5767750144, 0.7003160119, 0.8425520062, | |
60 1.0093879700, 1.1821349859, 1.3534560204, 1.5320819616, 1.7332619429, 1.9722349644, 2.3978140354, | |
61 -2.5756309032, -2.0573320389, -1.8984919786, -1.7727810144, -1.6662600040, -1.5742180347, -1.4993319511, | |
62 -1.4316639900, -1.3652280569, -1.3000990152, -1.2280930281, -1.1588579416, -1.0921250582, -1.0135740042, | |
63 -0.9202849865, -0.8287050128, -0.7374889851, -0.6447759867, -0.5590940118, -0.4857139885, -0.4110319912, | |
64 -0.3459700048, -0.2851159871, -0.2341620028, -0.1870580018, -0.1442500055, -0.1107169986, -0.0739680007, | |
65 -0.0365610011, -0.0073290002, 0.0203610007, 0.0479039997, 0.0751969963, 0.0980999991, 0.1220389977, | |
66 0.1458999962, 0.1694349945, 0.1970459968, 0.2252430022, 0.2556869984, 0.2870100141, 0.3197099864, | |
67 0.3525829911, 0.3889069855, 0.4334920049, 0.4769459963, 0.5204820037, 0.5644530058, 0.6122040153, | |
68 0.6685929894, 0.7341650128, 0.8032159805, 0.8784040213, 0.9566209912, 1.0397069454, 1.1293770075, | |
69 1.2211159468, 1.3080279827, 1.4024800062, 1.5056819916, 1.6227730513, 1.7724959850, 1.9430880547, | |
70 2.2903931141 | |
71 }; | |
72 | |
73 static const uint8_t nelly_band_sizes_table[NELLY_BANDS] = { | |
74 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 4, 4, 5, 6, 6, 7, 8, 9, 10, 12, 14, 15 | |
75 }; | |
76 | |
77 static const uint16_t nelly_init_table[64] = { | |
78 3134, 5342, 6870, 7792, 8569, 9185, 9744, 10191, 10631, 11061, 11434, 11770, | |
79 12116, 12513, 12925, 13300, 13674, 14027, 14352, 14716, 15117, 15477, 15824, | |
80 16157, 16513, 16804, 17090, 17401, 17679, 17948, 18238, 18520, 18764, 19078, | |
81 19381, 19640, 19921, 20205, 20500, 20813, 21162, 21465, 21794, 22137, 22453, | |
82 22756, 23067, 23350, 23636, 23926, 24227, 24521, 24819, 25107, 25414, 25730, | |
83 26120, 26497, 26895, 27344, 27877, 28463, 29426, 31355 | |
84 }; | |
85 | |
86 static const int16_t nelly_delta_table[32] = { | |
87 -11725, -9420, -7910, -6801, -5948, -5233, -4599, -4039, -3507, -3030, -2596, | |
88 -2170, -1774, -1383, -1016, -660, -329, -1, 337, 696, 1085, 1512, 1962, 2433, | |
89 2968, 3569, 4314, 5279, 6622, 8154, 10076, 12975 | |
90 }; | |
91 | |
92 typedef struct NellyMoserDecodeContext { | |
93 AVCodecContext* avctx; | |
94 float float_buf[NELLY_SAMPLES]; | |
95 float state[64]; | |
96 AVRandomState random_state; | |
97 GetBitContext gb; | |
98 int add_bias; | |
99 int scale_bias; | |
100 DSPContext dsp; | |
101 FFTContext fftc; | |
102 } NellyMoserDecodeContext; | |
103 | |
104 | |
105 DECLARE_ALIGNED_16(float,sine_window[128]); | |
106 DECLARE_ALIGNED_16(float,tcos[64]); | |
107 DECLARE_ALIGNED_16(float,tsin[64]); | |
108 DECLARE_ALIGNED_16(float,cos_tab[64]); | |
109 | |
110 static inline int signed_shift(int i, int shift) { | |
111 if (shift > 0) | |
112 return i << shift; | |
113 return i >> -shift; | |
114 } | |
115 | |
116 static void antialias(float *buf, float *audio) | |
117 { | |
118 int i, end, mid_hi, mid_lo; | |
119 | |
120 end = NELLY_BUF_LEN-1; | |
121 mid_hi = NELLY_BUF_LEN/2; | |
122 mid_lo = mid_hi-1; | |
123 | |
124 for (i = 0; i < NELLY_BUF_LEN/4; i++) { | |
125 audio[2*i] = buf[2*i ]*tcos[i ] - buf[end-2*i]*tsin[i]; | |
126 audio[2*i+1] = -(buf[end-2*i ]*tcos[i ] + buf[2*i ]*tsin[i]); | |
127 audio[end-2*i-1]= buf[end-2*i-1]*tcos[mid_lo-i] - buf[2*i+1 ]*tsin[mid_lo-i]; | |
128 audio[end-2*i ]= -(buf[2*i+1 ]*tcos[mid_lo-i] + buf[end-2*i]*tsin[mid_lo-i]); | |
129 } | |
130 } | |
131 | |
132 static void complex2signal(float *audio) | |
133 { | |
134 int i, end, mid_hi, mid_lo; | |
135 float *aptr, *sigptr, a, b, c, d, e, f, g; | |
136 | |
137 end = NELLY_BUF_LEN-1; | |
138 mid_hi = NELLY_BUF_LEN/2; | |
139 mid_lo = mid_hi-1; | |
140 | |
141 a = -audio[end]; | |
142 b = audio[end-1]; | |
143 c = -audio[1]; | |
144 d = cos_tab[0]; | |
145 e = audio[0]; | |
146 f = cos_tab[mid_lo]; | |
147 g = cos_tab[1]; | |
148 | |
149 audio[0] = d*e; | |
150 audio[1] = b*g-a*f; | |
151 audio[end-1] = a*g+b*f; | |
152 audio[end] = c*(-d); | |
153 | |
154 aptr = audio+end-2; | |
155 sigptr = cos_tab+mid_hi-1; | |
156 | |
157 for (i = 3; i < NELLY_BUF_LEN/2; i += 2) { | |
158 a = audio[i-1]; | |
159 b = -audio[i]; | |
160 c = cos_tab[i/2]; | |
161 d = *sigptr; | |
162 e = *(aptr-1); | |
163 f = -(*aptr); | |
164 | |
165 audio[i-1] = a*c+b*d; | |
166 *aptr = a*d-b*c; | |
167 | |
168 a = cos_tab[(i/2)+1]; | |
169 b = *(sigptr-1); | |
170 | |
171 *(aptr-1) = b*e+a*f; | |
172 audio[i] = a*e-b*f; | |
173 | |
174 sigptr--; | |
175 aptr -= 2; | |
176 } | |
177 } | |
178 | |
179 static void overlap_and_window(NellyMoserDecodeContext *s, float *state, float *audio) | |
180 { | |
181 int bot, mid_up, mid_down, top; | |
182 float s_bot, s_top; | |
183 | |
184 bot = 0; | |
185 top = NELLY_BUF_LEN-1; | |
186 mid_up = NELLY_BUF_LEN/2; | |
187 mid_down = (NELLY_BUF_LEN/2)-1; | |
188 | |
189 while (bot < NELLY_BUF_LEN/4) { | |
190 s_bot = audio[bot]; | |
191 s_top = audio[top]; | |
192 audio[bot] = (audio[mid_up]*sine_window[bot]-state[bot ]*sine_window[top])/s->scale_bias + s->add_bias; | |
193 audio[top] = (-state[bot ]*sine_window[bot]-audio[mid_up]*sine_window[top])/s->scale_bias + s->add_bias; | |
194 state[bot] = audio[mid_down]; | |
195 | |
196 audio[mid_down] = (s_top *sine_window[mid_down]-state[mid_down]*sine_window[mid_up])/s->scale_bias + s->add_bias; | |
197 audio[mid_up ] = (-state[mid_down]*sine_window[mid_down]-s_top *sine_window[mid_up])/s->scale_bias + s->add_bias; | |
198 state[mid_down] = s_bot; | |
199 | |
200 bot++; | |
201 mid_up++; | |
202 mid_down--; | |
203 top--; | |
204 } | |
205 } | |
206 | |
207 static int sum_bits(short *buf, short shift, short off) | |
208 { | |
209 int b, i = 0, ret = 0; | |
210 | |
211 for (i = 0; i < NELLY_FILL_LEN; i++) { | |
212 b = buf[i]-off; | |
213 b = ((b>>(shift-1))+1)>>1; | |
214 ret += av_clip(b, 0, NELLY_BIT_CAP); | |
215 } | |
216 | |
217 return ret; | |
218 } | |
219 | |
220 static int headroom(int *la) | |
221 { | |
222 int l; | |
223 if (*la == 0) { | |
224 return 31; | |
225 } | |
226 l = 30 - av_log2(FFABS(*la)); | |
227 *la <<= l; | |
228 return l; | |
229 } | |
230 | |
231 | |
232 static void get_sample_bits(float *buf, int *bits) | |
233 { | |
234 int i, j; | |
235 short sbuf[128]; | |
236 int bitsum = 0, last_bitsum, small_bitsum, big_bitsum; | |
237 short shift, shift_saved; | |
238 int max, sum, last_off, tmp; | |
239 int big_off, small_off; | |
240 int off; | |
241 | |
242 max = 0; | |
243 for (i = 0; i < NELLY_FILL_LEN; i++) { | |
244 max = FFMAX(max, buf[i]); | |
245 } | |
246 shift = -16; | |
247 shift += headroom(&max); | |
248 | |
249 sum = 0; | |
250 for (i = 0; i < NELLY_FILL_LEN; i++) { | |
251 sbuf[i] = signed_shift(buf[i], shift); | |
252 sbuf[i] = (3*sbuf[i])>>2; | |
253 sum += sbuf[i]; | |
254 } | |
255 | |
256 shift += 11; | |
257 shift_saved = shift; | |
258 sum -= NELLY_DETAIL_BITS << shift; | |
259 shift += headroom(&sum); | |
260 small_off = (NELLY_BASE_OFF * (sum>>16)) >> 15; | |
261 shift = shift_saved - (NELLY_BASE_SHIFT+shift-31); | |
262 | |
263 small_off = signed_shift(small_off, shift); | |
264 | |
265 bitsum = sum_bits(sbuf, shift_saved, small_off); | |
266 | |
267 if (bitsum != NELLY_DETAIL_BITS) { | |
268 shift = 0; | |
269 off = bitsum - NELLY_DETAIL_BITS; | |
270 | |
271 for(shift=0; FFABS(off) <= 16383; shift++) | |
272 off *= 2; | |
273 | |
274 off = (off * NELLY_BASE_OFF) >> 15; | |
275 shift = shift_saved-(NELLY_BASE_SHIFT+shift-15); | |
276 | |
277 off = signed_shift(off, shift); | |
278 | |
279 for (j = 1; j < 20; j++) { | |
280 last_off = small_off; | |
281 small_off += off; | |
282 last_bitsum = bitsum; | |
283 | |
284 bitsum = sum_bits(sbuf, shift_saved, small_off); | |
285 | |
286 if ((bitsum-NELLY_DETAIL_BITS) * (last_bitsum-NELLY_DETAIL_BITS) <= 0) | |
287 break; | |
288 } | |
289 | |
290 if (bitsum > NELLY_DETAIL_BITS) { | |
291 big_off = small_off; | |
292 small_off = last_off; | |
293 big_bitsum=bitsum; | |
294 small_bitsum=last_bitsum; | |
295 } else { | |
296 big_off = last_off; | |
297 big_bitsum=last_bitsum; | |
298 small_bitsum=bitsum; | |
299 } | |
300 | |
301 while (bitsum != NELLY_DETAIL_BITS && j <= 19) { | |
302 off = (big_off+small_off)>>1; | |
303 bitsum = sum_bits(sbuf, shift_saved, off); | |
304 if (bitsum > NELLY_DETAIL_BITS) { | |
305 big_off=off; | |
306 big_bitsum=bitsum; | |
307 } else { | |
308 small_off = off; | |
309 small_bitsum=bitsum; | |
310 } | |
311 j++; | |
312 } | |
313 | |
314 if (abs(big_bitsum-NELLY_DETAIL_BITS) >= | |
315 abs(small_bitsum-NELLY_DETAIL_BITS)) { | |
316 bitsum = small_bitsum; | |
317 } else { | |
318 small_off = big_off; | |
319 bitsum = big_bitsum; | |
320 } | |
321 } | |
322 | |
323 for (i = 0; i < NELLY_FILL_LEN; i++) { | |
324 tmp = sbuf[i]-small_off; | |
325 tmp = ((tmp>>(shift_saved-1))+1)>>1; | |
326 bits[i] = av_clip(tmp, 0, NELLY_BIT_CAP); | |
327 } | |
328 | |
329 if (bitsum > NELLY_DETAIL_BITS) { | |
330 tmp = i = 0; | |
331 while (tmp < NELLY_DETAIL_BITS) { | |
332 tmp += bits[i]; | |
333 i++; | |
334 } | |
335 | |
336 bits[i-1] -= tmp - NELLY_DETAIL_BITS; | |
337 for(; i < NELLY_FILL_LEN; i++) | |
338 bits[i] = 0; | |
339 } | |
340 } | |
341 | |
342 void nelly_decode_block(NellyMoserDecodeContext *s, unsigned char block[NELLY_BLOCK_LEN], float audio[NELLY_SAMPLES]) | |
343 { | |
344 int i,j; | |
345 float buf[NELLY_BUF_LEN], pows[NELLY_BUF_LEN]; | |
346 float *aptr, *bptr, *pptr, val, pval; | |
347 int bits[NELLY_BUF_LEN]; | |
348 unsigned char v; | |
349 | |
350 init_get_bits(&s->gb, block, NELLY_BLOCK_LEN * 8); | |
351 | |
352 bptr = buf; | |
353 pptr = pows; | |
354 val = nelly_init_table[get_bits(&s->gb, 6)]; | |
355 for (i=0 ; i<NELLY_BANDS ; i++) { | |
356 if (i > 0) | |
357 val += nelly_delta_table[get_bits(&s->gb, 5)]; | |
358 pval = pow(2, val/2048); | |
359 for (j = 0; j < nelly_band_sizes_table[i]; j++) { | |
360 *bptr++ = val; | |
361 *pptr++ = pval; | |
362 } | |
363 | |
364 } | |
365 | |
366 memset(&buf[NELLY_FILL_LEN],0,4*sizeof(float)); | |
367 memset(&pows[NELLY_FILL_LEN],0,4*sizeof(float)); | |
368 | |
369 get_sample_bits(buf, bits); | |
370 | |
371 for (i = 0; i < 2; i++) { | |
372 aptr = audio+i*128; | |
373 init_get_bits(&s->gb, block, NELLY_BLOCK_LEN * 8); | |
374 skip_bits(&s->gb, NELLY_HEADER_BITS + i*NELLY_DETAIL_BITS); | |
375 | |
376 for (j = 0; j < NELLY_FILL_LEN; j++) { | |
377 if (bits[j] <= 0) { | |
378 buf[j] = M_SQRT1_2*pows[j]; | |
379 if (av_random(&s->random_state) & 1) | |
380 buf[j] *= -1.0; | |
381 } else { | |
382 v = get_bits(&s->gb, bits[j]); | |
383 buf[j] = dequantization_table[(1<<bits[j])-1+v]*pows[j]; | |
384 } | |
385 } | |
386 | |
387 antialias(buf, aptr); | |
388 ff_fft_permute(&s->fftc, (FFTComplex*)aptr); | |
389 ff_fft_calc(&s->fftc, (FFTComplex*)aptr); | |
390 complex2signal(aptr); | |
391 overlap_and_window(s, s->state, aptr); | |
392 } | |
393 } | |
394 | |
395 static int decode_init(AVCodecContext * avctx) { | |
396 NellyMoserDecodeContext *s = avctx->priv_data; | |
397 int i; | |
398 float alpha; | |
399 | |
400 s->avctx = avctx; | |
401 av_init_random(0, &s->random_state); | |
402 ff_fft_init(&s->fftc, 6, 1); | |
403 dsputil_init(&s->dsp, avctx); | |
404 | |
405 if(s->dsp.float_to_int16 == ff_float_to_int16_c) { | |
406 s->add_bias = 385; | |
407 s->scale_bias = 32768; | |
408 } else { | |
409 s->add_bias = 0; | |
410 s->scale_bias = 1; | |
411 } | |
412 | |
413 /* Generate overlap window */ | |
414 if (!sine_window[0]) | |
415 for (i=0 ; i<128; i++) { | |
416 sine_window[i] = sin((i + 0.5) / 256.0 * M_PI); | |
417 } | |
418 | |
419 /* Generate tables */ | |
420 if (!tcos[0]) | |
421 for(i=0;i<64;i++) { | |
422 alpha = 2*M_PI * (i + 1.0 / 4.0) / 256; | |
423 tcos[i] = cos(alpha); | |
424 tsin[i] = -sin(alpha); | |
425 cos_tab[i] = cos(i/128.0*M_PI)/8.0; | |
426 } | |
427 | |
428 return 0; | |
429 } | |
430 | |
431 static int decode_tag(AVCodecContext * avctx, | |
432 void *data, int *data_size, | |
433 uint8_t * buf, int buf_size) { | |
434 NellyMoserDecodeContext *s = avctx->priv_data; | |
435 int blocks, i; | |
436 int16_t* samples; | |
437 *data_size = 0; | |
438 samples = (int16_t*)data; | |
439 | |
440 if (buf_size < avctx->block_align) | |
441 return buf_size; | |
442 | |
443 switch (buf_size) { | |
444 case 64: // 8000Hz | |
445 blocks = 1; break; | |
446 case 128: // 11025Hz | |
447 blocks = 2; break; | |
448 case 256: // 22050Hz | |
449 blocks = 4; break; | |
450 default: | |
451 av_log(avctx, AV_LOG_DEBUG, "Tag size %d unknown, report sample!\n", buf_size); | |
452 return buf_size; | |
453 } | |
454 | |
455 for (i=0 ; i<blocks ; i++) { | |
456 nelly_decode_block(s, &buf[i*NELLY_BLOCK_LEN], s->float_buf); | |
457 s->dsp.float_to_int16(&samples[i*NELLY_SAMPLES], s->float_buf, NELLY_SAMPLES); | |
458 *data_size += NELLY_SAMPLES*sizeof(int16_t); | |
459 } | |
460 | |
461 return blocks*NELLY_SAMPLES*sizeof(int16_t); | |
462 } | |
463 | |
464 static int decode_end(AVCodecContext * avctx) { | |
465 NellyMoserDecodeContext *s = avctx->priv_data; | |
466 | |
467 ff_fft_end(&s->fftc); | |
468 return 0; | |
469 } | |
470 | |
471 AVCodec nellymoser_decoder = { | |
472 "nellymoser", | |
473 CODEC_TYPE_AUDIO, | |
474 CODEC_ID_NELLYMOSER, | |
475 sizeof(NellyMoserDecodeContext), | |
476 decode_init, | |
477 NULL, | |
478 decode_end, | |
479 decode_tag, | |
480 }; | |
481 |