Change the "report sample" message from AV_LOG_DEBUG to AV_LOG_ERROR.
[libav.git] / libavcodec / nellymoserdec.c
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 (2 * NELLY_BUF_LEN)
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 DECLARE_ALIGNED_16(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 MDCTContext imdct_ctx;
102 DECLARE_ALIGNED_16(float,imdct_tmp[NELLY_BUF_LEN]);
103 DECLARE_ALIGNED_16(float,imdct_out[NELLY_BUF_LEN * 2]);
104 } NellyMoserDecodeContext;
105
106 static DECLARE_ALIGNED_16(float,sine_window[128]);
107
108 static inline int signed_shift(int i, int shift) {
109 if (shift > 0)
110 return i << shift;
111 return i >> -shift;
112 }
113
114 static void overlap_and_window(NellyMoserDecodeContext *s, float *state, float *audio)
115 {
116 int bot, mid_up, mid_down, top;
117 float s_bot, s_top;
118
119 bot = 0;
120 top = NELLY_BUF_LEN-1;
121 mid_up = NELLY_BUF_LEN/2;
122 mid_down = (NELLY_BUF_LEN/2)-1;
123
124 while (bot < NELLY_BUF_LEN/4) {
125 s_bot = audio[bot];
126 s_top = audio[top];
127 audio[bot] = (audio[mid_up]*sine_window[bot]-state[bot ]*sine_window[top])/s->scale_bias + s->add_bias;
128 audio[top] = (-state[bot ]*sine_window[bot]-audio[mid_up]*sine_window[top])/s->scale_bias + s->add_bias;
129 state[bot] = audio[mid_down];
130
131 audio[mid_down] = (s_top *sine_window[mid_down]-state[mid_down]*sine_window[mid_up])/s->scale_bias + s->add_bias;
132 audio[mid_up ] = (-state[mid_down]*sine_window[mid_down]-s_top *sine_window[mid_up])/s->scale_bias + s->add_bias;
133 state[mid_down] = s_bot;
134
135 bot++;
136 mid_up++;
137 mid_down--;
138 top--;
139 }
140 }
141
142 static int sum_bits(short *buf, short shift, short off)
143 {
144 int b, i = 0, ret = 0;
145
146 for (i = 0; i < NELLY_FILL_LEN; i++) {
147 b = buf[i]-off;
148 b = ((b>>(shift-1))+1)>>1;
149 ret += av_clip(b, 0, NELLY_BIT_CAP);
150 }
151
152 return ret;
153 }
154
155 static int headroom(int *la)
156 {
157 int l;
158 if (*la == 0) {
159 return 31;
160 }
161 l = 30 - av_log2(FFABS(*la));
162 *la <<= l;
163 return l;
164 }
165
166
167 static void get_sample_bits(const float *buf, int *bits)
168 {
169 int i, j;
170 short sbuf[128];
171 int bitsum = 0, last_bitsum, small_bitsum, big_bitsum;
172 short shift, shift_saved;
173 int max, sum, last_off, tmp;
174 int big_off, small_off;
175 int off;
176
177 max = 0;
178 for (i = 0; i < NELLY_FILL_LEN; i++) {
179 max = FFMAX(max, buf[i]);
180 }
181 shift = -16;
182 shift += headroom(&max);
183
184 sum = 0;
185 for (i = 0; i < NELLY_FILL_LEN; i++) {
186 sbuf[i] = signed_shift(buf[i], shift);
187 sbuf[i] = (3*sbuf[i])>>2;
188 sum += sbuf[i];
189 }
190
191 shift += 11;
192 shift_saved = shift;
193 sum -= NELLY_DETAIL_BITS << shift;
194 shift += headroom(&sum);
195 small_off = (NELLY_BASE_OFF * (sum>>16)) >> 15;
196 shift = shift_saved - (NELLY_BASE_SHIFT+shift-31);
197
198 small_off = signed_shift(small_off, shift);
199
200 bitsum = sum_bits(sbuf, shift_saved, small_off);
201
202 if (bitsum != NELLY_DETAIL_BITS) {
203 shift = 0;
204 off = bitsum - NELLY_DETAIL_BITS;
205
206 for(shift=0; FFABS(off) <= 16383; shift++)
207 off *= 2;
208
209 off = (off * NELLY_BASE_OFF) >> 15;
210 shift = shift_saved-(NELLY_BASE_SHIFT+shift-15);
211
212 off = signed_shift(off, shift);
213
214 for (j = 1; j < 20; j++) {
215 last_off = small_off;
216 small_off += off;
217 last_bitsum = bitsum;
218
219 bitsum = sum_bits(sbuf, shift_saved, small_off);
220
221 if ((bitsum-NELLY_DETAIL_BITS) * (last_bitsum-NELLY_DETAIL_BITS) <= 0)
222 break;
223 }
224
225 if (bitsum > NELLY_DETAIL_BITS) {
226 big_off = small_off;
227 small_off = last_off;
228 big_bitsum=bitsum;
229 small_bitsum=last_bitsum;
230 } else {
231 big_off = last_off;
232 big_bitsum=last_bitsum;
233 small_bitsum=bitsum;
234 }
235
236 while (bitsum != NELLY_DETAIL_BITS && j <= 19) {
237 off = (big_off+small_off)>>1;
238 bitsum = sum_bits(sbuf, shift_saved, off);
239 if (bitsum > NELLY_DETAIL_BITS) {
240 big_off=off;
241 big_bitsum=bitsum;
242 } else {
243 small_off = off;
244 small_bitsum=bitsum;
245 }
246 j++;
247 }
248
249 if (abs(big_bitsum-NELLY_DETAIL_BITS) >=
250 abs(small_bitsum-NELLY_DETAIL_BITS)) {
251 bitsum = small_bitsum;
252 } else {
253 small_off = big_off;
254 bitsum = big_bitsum;
255 }
256 }
257
258 for (i = 0; i < NELLY_FILL_LEN; i++) {
259 tmp = sbuf[i]-small_off;
260 tmp = ((tmp>>(shift_saved-1))+1)>>1;
261 bits[i] = av_clip(tmp, 0, NELLY_BIT_CAP);
262 }
263
264 if (bitsum > NELLY_DETAIL_BITS) {
265 tmp = i = 0;
266 while (tmp < NELLY_DETAIL_BITS) {
267 tmp += bits[i];
268 i++;
269 }
270
271 bits[i-1] -= tmp - NELLY_DETAIL_BITS;
272 for(; i < NELLY_FILL_LEN; i++)
273 bits[i] = 0;
274 }
275 }
276
277 void nelly_decode_block(NellyMoserDecodeContext *s, unsigned char block[NELLY_BLOCK_LEN], float audio[NELLY_SAMPLES])
278 {
279 int i,j;
280 float buf[NELLY_FILL_LEN], pows[NELLY_FILL_LEN];
281 float *aptr, *bptr, *pptr, val, pval;
282 int bits[NELLY_BUF_LEN];
283 unsigned char v;
284 float a;
285
286 init_get_bits(&s->gb, block, NELLY_BLOCK_LEN * 8);
287
288 bptr = buf;
289 pptr = pows;
290 val = nelly_init_table[get_bits(&s->gb, 6)];
291 for (i=0 ; i<NELLY_BANDS ; i++) {
292 if (i > 0)
293 val += nelly_delta_table[get_bits(&s->gb, 5)];
294 pval = pow(2, val/2048);
295 for (j = 0; j < nelly_band_sizes_table[i]; j++) {
296 *bptr++ = val;
297 *pptr++ = pval;
298 }
299
300 }
301
302 get_sample_bits(buf, bits);
303
304 for (i = 0; i < 2; i++) {
305 aptr = audio + i * NELLY_BUF_LEN;
306
307 init_get_bits(&s->gb, block, NELLY_BLOCK_LEN * 8);
308 skip_bits(&s->gb, NELLY_HEADER_BITS + i*NELLY_DETAIL_BITS);
309
310 for (j = 0; j < NELLY_FILL_LEN; j++) {
311 if (bits[j] <= 0) {
312 aptr[j] = M_SQRT1_2*pows[j];
313 if (av_random(&s->random_state) & 1)
314 aptr[j] *= -1.0;
315 } else {
316 v = get_bits(&s->gb, bits[j]);
317 aptr[j] = dequantization_table[(1<<bits[j])-1+v]*pows[j];
318 }
319 }
320 memset(&aptr[NELLY_FILL_LEN], 0,
321 (NELLY_BUF_LEN - NELLY_FILL_LEN) * sizeof(float));
322
323 s->imdct_ctx.fft.imdct_calc(&s->imdct_ctx, s->imdct_out,
324 aptr, s->imdct_tmp);
325 /* XXX: overlapping and windowing should be part of a more
326 generic imdct function */
327 a = 1.0 / 8.0;
328 for(j = 0; j < NELLY_BUF_LEN / 2; j++) {
329 aptr[j] = s->imdct_out[j + NELLY_BUF_LEN + NELLY_BUF_LEN / 2] * a;
330 aptr[j + NELLY_BUF_LEN / 2] = -s->imdct_out[j] * a;
331 }
332 overlap_and_window(s, s->state, aptr);
333 }
334 }
335
336 static int decode_init(AVCodecContext * avctx) {
337 NellyMoserDecodeContext *s = avctx->priv_data;
338 int i;
339
340 s->avctx = avctx;
341 av_init_random(0, &s->random_state);
342 ff_mdct_init(&s->imdct_ctx, 8, 1);
343
344 dsputil_init(&s->dsp, avctx);
345
346 if(s->dsp.float_to_int16 == ff_float_to_int16_c) {
347 s->add_bias = 385;
348 s->scale_bias = 32768;
349 } else {
350 s->add_bias = 0;
351 s->scale_bias = 1;
352 }
353
354 /* Generate overlap window */
355 if (!sine_window[0])
356 for (i=0 ; i<128; i++) {
357 sine_window[i] = sin((i + 0.5) / 256.0 * M_PI);
358 }
359
360 return 0;
361 }
362
363 static int decode_tag(AVCodecContext * avctx,
364 void *data, int *data_size,
365 uint8_t * buf, int buf_size) {
366 NellyMoserDecodeContext *s = avctx->priv_data;
367 int blocks, i;
368 int16_t* samples;
369 *data_size = 0;
370 samples = (int16_t*)data;
371
372 if (buf_size < avctx->block_align)
373 return buf_size;
374
375 switch (buf_size) {
376 case 64: // 8000Hz
377 blocks = 1; break;
378 case 128: // 11025Hz
379 blocks = 2; break;
380 case 256: // 22050Hz
381 blocks = 4; break;
382 default:
383 av_log(avctx, AV_LOG_ERROR, "Tag size %d unknown, report sample!\n", buf_size);
384 return buf_size;
385 }
386
387 for (i=0 ; i<blocks ; i++) {
388 nelly_decode_block(s, &buf[i*NELLY_BLOCK_LEN], s->float_buf);
389 s->dsp.float_to_int16(&samples[i*NELLY_SAMPLES], s->float_buf, NELLY_SAMPLES);
390 *data_size += NELLY_SAMPLES*sizeof(int16_t);
391 }
392
393 return buf_size;
394 }
395
396 static int decode_end(AVCodecContext * avctx) {
397 NellyMoserDecodeContext *s = avctx->priv_data;
398
399 ff_mdct_end(&s->imdct_ctx);
400 return 0;
401 }
402
403 AVCodec nellymoser_decoder = {
404 "nellymoser",
405 CODEC_TYPE_AUDIO,
406 CODEC_ID_NELLYMOSER,
407 sizeof(NellyMoserDecodeContext),
408 decode_init,
409 NULL,
410 decode_end,
411 decode_tag,
412 };
413