2 * Windows Media Audio Voice decoder.
3 * Copyright (c) 2009 Ronald S. Bultje
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * @brief Windows Media Audio Voice compatible decoder
25 * @author Ronald S. Bultje <rsbultje@gmail.com>
28 #define UNCHECKED_BITSTREAM_READER 1
32 #include "libavutil/channel_layout.h"
33 #include "libavutil/mem.h"
38 #include "wmavoice_data.h"
39 #include "celp_filters.h"
40 #include "acelp_vectors.h"
41 #include "acelp_filters.h"
47 #define MAX_BLOCKS 8 ///< maximum number of blocks per frame
48 #define MAX_LSPS 16 ///< maximum filter order
49 #define MAX_LSPS_ALIGN16 16 ///< same as #MAX_LSPS; needs to be multiple
50 ///< of 16 for ASM input buffer alignment
51 #define MAX_FRAMES 3 ///< maximum number of frames per superframe
52 #define MAX_FRAMESIZE 160 ///< maximum number of samples per frame
53 #define MAX_SIGNAL_HISTORY 416 ///< maximum excitation signal history
54 #define MAX_SFRAMESIZE (MAX_FRAMESIZE * MAX_FRAMES)
55 ///< maximum number of samples per superframe
56 #define SFRAME_CACHE_MAXSIZE 256 ///< maximum cache size for frame data that
57 ///< was split over two packets
58 #define VLC_NBITS 6 ///< number of bits to read per VLC iteration
61 * Frame type VLC coding.
63 static VLC frame_type_vlc;
66 * Adaptive codebook types.
69 ACB_TYPE_NONE = 0, ///< no adaptive codebook (only hardcoded fixed)
70 ACB_TYPE_ASYMMETRIC = 1, ///< adaptive codebook with per-frame pitch, which
71 ///< we interpolate to get a per-sample pitch.
72 ///< Signal is generated using an asymmetric sinc
74 ///< @note see #wmavoice_ipol1_coeffs
75 ACB_TYPE_HAMMING = 2 ///< Per-block pitch with signal generation using
76 ///< a Hamming sinc window function
77 ///< @note see #wmavoice_ipol2_coeffs
81 * Fixed codebook types.
84 FCB_TYPE_SILENCE = 0, ///< comfort noise during silence
85 ///< generated from a hardcoded (fixed) codebook
86 ///< with per-frame (low) gain values
87 FCB_TYPE_HARDCODED = 1, ///< hardcoded (fixed) codebook with per-block
89 FCB_TYPE_AW_PULSES = 2, ///< Pitch-adaptive window (AW) pulse signals,
90 ///< used in particular for low-bitrate streams
91 FCB_TYPE_EXC_PULSES = 3, ///< Innovation (fixed) codebook pulse sets in
92 ///< combinations of either single pulses or
97 * Description of frame types.
99 static const struct frame_type_desc {
100 uint8_t n_blocks; ///< amount of blocks per frame (each block
101 ///< (contains 160/#n_blocks samples)
102 uint8_t log_n_blocks; ///< log2(#n_blocks)
103 uint8_t acb_type; ///< Adaptive codebook type (ACB_TYPE_*)
104 uint8_t fcb_type; ///< Fixed codebook type (FCB_TYPE_*)
105 uint8_t dbl_pulses; ///< how many pulse vectors have pulse pairs
106 ///< (rather than just one single pulse)
107 ///< only if #fcb_type == #FCB_TYPE_EXC_PULSES
108 uint16_t frame_size; ///< the amount of bits that make up the block
109 ///< data (per frame)
110 } frame_descs[17] = {
111 { 1, 0, ACB_TYPE_NONE, FCB_TYPE_SILENCE, 0, 0 },
112 { 2, 1, ACB_TYPE_NONE, FCB_TYPE_HARDCODED, 0, 28 },
113 { 2, 1, ACB_TYPE_ASYMMETRIC, FCB_TYPE_AW_PULSES, 0, 46 },
114 { 2, 1, ACB_TYPE_ASYMMETRIC, FCB_TYPE_EXC_PULSES, 2, 80 },
115 { 2, 1, ACB_TYPE_ASYMMETRIC, FCB_TYPE_EXC_PULSES, 5, 104 },
116 { 4, 2, ACB_TYPE_ASYMMETRIC, FCB_TYPE_EXC_PULSES, 0, 108 },
117 { 4, 2, ACB_TYPE_ASYMMETRIC, FCB_TYPE_EXC_PULSES, 2, 132 },
118 { 4, 2, ACB_TYPE_ASYMMETRIC, FCB_TYPE_EXC_PULSES, 5, 168 },
119 { 2, 1, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 0, 64 },
120 { 2, 1, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 2, 80 },
121 { 2, 1, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 5, 104 },
122 { 4, 2, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 0, 108 },
123 { 4, 2, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 2, 132 },
124 { 4, 2, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 5, 168 },
125 { 8, 3, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 0, 176 },
126 { 8, 3, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 2, 208 },
127 { 8, 3, ACB_TYPE_HAMMING, FCB_TYPE_EXC_PULSES, 5, 256 }
131 * WMA Voice decoding context.
135 * @name Global values specified in the stream header / extradata or used all over.
139 GetBitContext gb; ///< packet bitreader. During decoder init,
140 ///< it contains the extradata from the
141 ///< demuxer. During decoding, it contains
143 int8_t vbm_tree[25]; ///< converts VLC codes to frame type
145 int spillover_bitsize; ///< number of bits used to specify
146 ///< #spillover_nbits in the packet header
147 ///< = ceil(log2(ctx->block_align << 3))
148 int history_nsamples; ///< number of samples in history for signal
149 ///< prediction (through ACB)
151 /* postfilter specific values */
152 int do_apf; ///< whether to apply the averaged
153 ///< projection filter (APF)
154 int denoise_strength; ///< strength of denoising in Wiener filter
156 int denoise_tilt_corr; ///< Whether to apply tilt correction to the
157 ///< Wiener filter coefficients (postfilter)
158 int dc_level; ///< Predicted amount of DC noise, based
159 ///< on which a DC removal filter is used
161 int lsps; ///< number of LSPs per frame [10 or 16]
162 int lsp_q_mode; ///< defines quantizer defaults [0, 1]
163 int lsp_def_mode; ///< defines different sets of LSP defaults
165 int frame_lsp_bitsize; ///< size (in bits) of LSPs, when encoded
166 ///< per-frame (independent coding)
167 int sframe_lsp_bitsize; ///< size (in bits) of LSPs, when encoded
168 ///< per superframe (residual coding)
170 int min_pitch_val; ///< base value for pitch parsing code
171 int max_pitch_val; ///< max value + 1 for pitch parsing
172 int pitch_nbits; ///< number of bits used to specify the
173 ///< pitch value in the frame header
174 int block_pitch_nbits; ///< number of bits used to specify the
175 ///< first block's pitch value
176 int block_pitch_range; ///< range of the block pitch
177 int block_delta_pitch_nbits; ///< number of bits used to specify the
178 ///< delta pitch between this and the last
179 ///< block's pitch value, used in all but
181 int block_delta_pitch_hrange; ///< 1/2 range of the delta (full range is
182 ///< from -this to +this-1)
183 uint16_t block_conv_table[4]; ///< boundaries for block pitch unit/scale
189 * @name Packet values specified in the packet header or related to a packet.
191 * A packet is considered to be a single unit of data provided to this
192 * decoder by the demuxer.
195 int spillover_nbits; ///< number of bits of the previous packet's
196 ///< last superframe preceding this
197 ///< packet's first full superframe (useful
198 ///< for re-synchronization also)
199 int has_residual_lsps; ///< if set, superframes contain one set of
200 ///< LSPs that cover all frames, encoded as
201 ///< independent and residual LSPs; if not
202 ///< set, each frame contains its own, fully
203 ///< independent, LSPs
204 int skip_bits_next; ///< number of bits to skip at the next call
205 ///< to #wmavoice_decode_packet() (since
206 ///< they're part of the previous superframe)
208 uint8_t sframe_cache[SFRAME_CACHE_MAXSIZE + FF_INPUT_BUFFER_PADDING_SIZE];
209 ///< cache for superframe data split over
210 ///< multiple packets
211 int sframe_cache_size; ///< set to >0 if we have data from an
212 ///< (incomplete) superframe from a previous
213 ///< packet that spilled over in the current
214 ///< packet; specifies the amount of bits in
216 PutBitContext pb; ///< bitstream writer for #sframe_cache
221 * @name Frame and superframe values
222 * Superframe and frame data - these can change from frame to frame,
223 * although some of them do in that case serve as a cache / history for
224 * the next frame or superframe.
227 double prev_lsps[MAX_LSPS]; ///< LSPs of the last frame of the previous
229 int last_pitch_val; ///< pitch value of the previous frame
230 int last_acb_type; ///< frame type [0-2] of the previous frame
231 int pitch_diff_sh16; ///< ((cur_pitch_val - #last_pitch_val)
232 ///< << 16) / #MAX_FRAMESIZE
233 float silence_gain; ///< set for use in blocks if #ACB_TYPE_NONE
235 int aw_idx_is_ext; ///< whether the AW index was encoded in
236 ///< 8 bits (instead of 6)
237 int aw_pulse_range; ///< the range over which #aw_pulse_set1()
238 ///< can apply the pulse, relative to the
239 ///< value in aw_first_pulse_off. The exact
240 ///< position of the first AW-pulse is within
241 ///< [pulse_off, pulse_off + this], and
242 ///< depends on bitstream values; [16 or 24]
243 int aw_n_pulses[2]; ///< number of AW-pulses in each block; note
244 ///< that this number can be negative (in
245 ///< which case it basically means "zero")
246 int aw_first_pulse_off[2]; ///< index of first sample to which to
247 ///< apply AW-pulses, or -0xff if unset
248 int aw_next_pulse_off_cache; ///< the position (relative to start of the
249 ///< second block) at which pulses should
250 ///< start to be positioned, serves as a
251 ///< cache for pitch-adaptive window pulses
254 int frame_cntr; ///< current frame index [0 - 0xFFFE]; is
255 ///< only used for comfort noise in #pRNG()
256 float gain_pred_err[6]; ///< cache for gain prediction
257 float excitation_history[MAX_SIGNAL_HISTORY];
258 ///< cache of the signal of previous
259 ///< superframes, used as a history for
260 ///< signal generation
261 float synth_history[MAX_LSPS]; ///< see #excitation_history
265 * @name Postfilter values
267 * Variables used for postfilter implementation, mostly history for
268 * smoothing and so on, and context variables for FFT/iFFT.
271 RDFTContext rdft, irdft; ///< contexts for FFT-calculation in the
272 ///< postfilter (for denoise filter)
273 DCTContext dct, dst; ///< contexts for phase shift (in Hilbert
274 ///< transform, part of postfilter)
275 float sin[511], cos[511]; ///< 8-bit cosine/sine windows over [-pi,pi]
277 float postfilter_agc; ///< gain control memory, used in
278 ///< #adaptive_gain_control()
279 float dcf_mem[2]; ///< DC filter history
280 float zero_exc_pf[MAX_SIGNAL_HISTORY + MAX_SFRAMESIZE];
281 ///< zero filter output (i.e. excitation)
283 float denoise_filter_cache[MAX_FRAMESIZE];
284 int denoise_filter_cache_size; ///< samples in #denoise_filter_cache
285 DECLARE_ALIGNED(32, float, tilted_lpcs_pf)[0x80];
286 ///< aligned buffer for LPC tilting
287 DECLARE_ALIGNED(32, float, denoise_coeffs_pf)[0x80];
288 ///< aligned buffer for denoise coefficients
289 DECLARE_ALIGNED(32, float, synth_filter_out_buf)[0x80 + MAX_LSPS_ALIGN16];
290 ///< aligned buffer for postfilter speech
298 * Set up the variable bit mode (VBM) tree from container extradata.
299 * @param gb bit I/O context.
300 * The bit context (s->gb) should be loaded with byte 23-46 of the
301 * container extradata (i.e. the ones containing the VBM tree).
302 * @param vbm_tree pointer to array to which the decoded VBM tree will be
304 * @return 0 on success, <0 on error.
306 static av_cold int decode_vbmtree(GetBitContext *gb, int8_t vbm_tree[25])
308 static const uint8_t bits[] = {
311 10, 10, 10, 12, 12, 12,
314 static const uint16_t codes[] = {
315 0x0000, 0x0001, 0x0002, // 00/01/10
316 0x000c, 0x000d, 0x000e, // 11+00/01/10
317 0x003c, 0x003d, 0x003e, // 1111+00/01/10
318 0x00fc, 0x00fd, 0x00fe, // 111111+00/01/10
319 0x03fc, 0x03fd, 0x03fe, // 11111111+00/01/10
320 0x0ffc, 0x0ffd, 0x0ffe, // 1111111111+00/01/10
321 0x3ffc, 0x3ffd, 0x3ffe, 0x3fff // 111111111111+xx
323 int cntr[8] = { 0 }, n, res;
325 memset(vbm_tree, 0xff, sizeof(vbm_tree[0]) * 25);
326 for (n = 0; n < 17; n++) {
327 res = get_bits(gb, 3);
328 if (cntr[res] > 3) // should be >= 3 + (res == 7))
330 vbm_tree[res * 3 + cntr[res]++] = n;
332 INIT_VLC_STATIC(&frame_type_vlc, VLC_NBITS, sizeof(bits),
333 bits, 1, 1, codes, 2, 2, 132);
338 * Set up decoder with parameters from demuxer (extradata etc.).
340 static av_cold int wmavoice_decode_init(AVCodecContext *ctx)
342 int n, flags, pitch_range, lsp16_flag;
343 WMAVoiceContext *s = ctx->priv_data;
347 * - byte 0-18: WMAPro-in-WMAVoice extradata (see wmaprodec.c),
348 * - byte 19-22: flags field (annoyingly in LE; see below for known
350 * - byte 23-46: variable bitmode tree (really just 17 * 3 bits,
353 if (ctx->extradata_size != 46) {
354 av_log(ctx, AV_LOG_ERROR,
355 "Invalid extradata size %d (should be 46)\n",
356 ctx->extradata_size);
359 flags = AV_RL32(ctx->extradata + 18);
360 s->spillover_bitsize = 3 + av_ceil_log2(ctx->block_align);
361 s->do_apf = flags & 0x1;
363 ff_rdft_init(&s->rdft, 7, DFT_R2C);
364 ff_rdft_init(&s->irdft, 7, IDFT_C2R);
365 ff_dct_init(&s->dct, 6, DCT_I);
366 ff_dct_init(&s->dst, 6, DST_I);
368 ff_sine_window_init(s->cos, 256);
369 memcpy(&s->sin[255], s->cos, 256 * sizeof(s->cos[0]));
370 for (n = 0; n < 255; n++) {
371 s->sin[n] = -s->sin[510 - n];
372 s->cos[510 - n] = s->cos[n];
375 s->denoise_strength = (flags >> 2) & 0xF;
376 if (s->denoise_strength >= 12) {
377 av_log(ctx, AV_LOG_ERROR,
378 "Invalid denoise filter strength %d (max=11)\n",
379 s->denoise_strength);
382 s->denoise_tilt_corr = !!(flags & 0x40);
383 s->dc_level = (flags >> 7) & 0xF;
384 s->lsp_q_mode = !!(flags & 0x2000);
385 s->lsp_def_mode = !!(flags & 0x4000);
386 lsp16_flag = flags & 0x1000;
389 s->frame_lsp_bitsize = 34;
390 s->sframe_lsp_bitsize = 60;
393 s->frame_lsp_bitsize = 24;
394 s->sframe_lsp_bitsize = 48;
396 for (n = 0; n < s->lsps; n++)
397 s->prev_lsps[n] = M_PI * (n + 1.0) / (s->lsps + 1.0);
399 init_get_bits(&s->gb, ctx->extradata + 22, (ctx->extradata_size - 22) << 3);
400 if (decode_vbmtree(&s->gb, s->vbm_tree) < 0) {
401 av_log(ctx, AV_LOG_ERROR, "Invalid VBM tree; broken extradata?\n");
405 s->min_pitch_val = ((ctx->sample_rate << 8) / 400 + 50) >> 8;
406 s->max_pitch_val = ((ctx->sample_rate << 8) * 37 / 2000 + 50) >> 8;
407 pitch_range = s->max_pitch_val - s->min_pitch_val;
408 if (pitch_range <= 0) {
409 av_log(ctx, AV_LOG_ERROR, "Invalid pitch range; broken extradata?\n");
412 s->pitch_nbits = av_ceil_log2(pitch_range);
413 s->last_pitch_val = 40;
414 s->last_acb_type = ACB_TYPE_NONE;
415 s->history_nsamples = s->max_pitch_val + 8;
417 if (s->min_pitch_val < 1 || s->history_nsamples > MAX_SIGNAL_HISTORY) {
418 int min_sr = ((((1 << 8) - 50) * 400) + 0xFF) >> 8,
419 max_sr = ((((MAX_SIGNAL_HISTORY - 8) << 8) + 205) * 2000 / 37) >> 8;
421 av_log(ctx, AV_LOG_ERROR,
422 "Unsupported samplerate %d (min=%d, max=%d)\n",
423 ctx->sample_rate, min_sr, max_sr); // 322-22097 Hz
428 s->block_conv_table[0] = s->min_pitch_val;
429 s->block_conv_table[1] = (pitch_range * 25) >> 6;
430 s->block_conv_table[2] = (pitch_range * 44) >> 6;
431 s->block_conv_table[3] = s->max_pitch_val - 1;
432 s->block_delta_pitch_hrange = (pitch_range >> 3) & ~0xF;
433 if (s->block_delta_pitch_hrange <= 0) {
434 av_log(ctx, AV_LOG_ERROR, "Invalid delta pitch hrange; broken extradata?\n");
437 s->block_delta_pitch_nbits = 1 + av_ceil_log2(s->block_delta_pitch_hrange);
438 s->block_pitch_range = s->block_conv_table[2] +
439 s->block_conv_table[3] + 1 +
440 2 * (s->block_conv_table[1] - 2 * s->min_pitch_val);
441 s->block_pitch_nbits = av_ceil_log2(s->block_pitch_range);
444 ctx->channel_layout = AV_CH_LAYOUT_MONO;
445 ctx->sample_fmt = AV_SAMPLE_FMT_FLT;
447 avcodec_get_frame_defaults(&s->frame);
448 ctx->coded_frame = &s->frame;
454 * @name Postfilter functions
455 * Postfilter functions (gain control, wiener denoise filter, DC filter,
456 * kalman smoothening, plus surrounding code to wrap it)
460 * Adaptive gain control (as used in postfilter).
462 * Identical to #ff_adaptive_gain_control() in acelp_vectors.c, except
463 * that the energy here is calculated using sum(abs(...)), whereas the
464 * other codecs (e.g. AMR-NB, SIPRO) use sqrt(dotproduct(...)).
466 * @param out output buffer for filtered samples
467 * @param in input buffer containing the samples as they are after the
468 * postfilter steps so far
469 * @param speech_synth input buffer containing speech synth before postfilter
470 * @param size input buffer size
471 * @param alpha exponential filter factor
472 * @param gain_mem pointer to filter memory (single float)
474 static void adaptive_gain_control(float *out, const float *in,
475 const float *speech_synth,
476 int size, float alpha, float *gain_mem)
479 float speech_energy = 0.0, postfilter_energy = 0.0, gain_scale_factor;
480 float mem = *gain_mem;
482 for (i = 0; i < size; i++) {
483 speech_energy += fabsf(speech_synth[i]);
484 postfilter_energy += fabsf(in[i]);
486 gain_scale_factor = (1.0 - alpha) * speech_energy / postfilter_energy;
488 for (i = 0; i < size; i++) {
489 mem = alpha * mem + gain_scale_factor;
490 out[i] = in[i] * mem;
497 * Kalman smoothing function.
499 * This function looks back pitch +/- 3 samples back into history to find
500 * the best fitting curve (that one giving the optimal gain of the two
501 * signals, i.e. the highest dot product between the two), and then
502 * uses that signal history to smoothen the output of the speech synthesis
505 * @param s WMA Voice decoding context
506 * @param pitch pitch of the speech signal
507 * @param in input speech signal
508 * @param out output pointer for smoothened signal
509 * @param size input/output buffer size
511 * @returns -1 if no smoothening took place, e.g. because no optimal
512 * fit could be found, or 0 on success.
514 static int kalman_smoothen(WMAVoiceContext *s, int pitch,
515 const float *in, float *out, int size)
518 float optimal_gain = 0, dot;
519 const float *ptr = &in[-FFMAX(s->min_pitch_val, pitch - 3)],
520 *end = &in[-FFMIN(s->max_pitch_val, pitch + 3)],
521 *best_hist_ptr = NULL;
523 /* find best fitting point in history */
525 dot = ff_scalarproduct_float_c(in, ptr, size);
526 if (dot > optimal_gain) {
530 } while (--ptr >= end);
532 if (optimal_gain <= 0)
534 dot = ff_scalarproduct_float_c(best_hist_ptr, best_hist_ptr, size);
535 if (dot <= 0) // would be 1.0
538 if (optimal_gain <= dot) {
539 dot = dot / (dot + 0.6 * optimal_gain); // 0.625-1.000
543 /* actual smoothing */
544 for (n = 0; n < size; n++)
545 out[n] = best_hist_ptr[n] + dot * (in[n] - best_hist_ptr[n]);
551 * Get the tilt factor of a formant filter from its transfer function
552 * @see #tilt_factor() in amrnbdec.c, which does essentially the same,
553 * but somehow (??) it does a speech synthesis filter in the
554 * middle, which is missing here
556 * @param lpcs LPC coefficients
557 * @param n_lpcs Size of LPC buffer
558 * @returns the tilt factor
560 static float tilt_factor(const float *lpcs, int n_lpcs)
564 rh0 = 1.0 + ff_scalarproduct_float_c(lpcs, lpcs, n_lpcs);
565 rh1 = lpcs[0] + ff_scalarproduct_float_c(lpcs, &lpcs[1], n_lpcs - 1);
571 * Derive denoise filter coefficients (in real domain) from the LPCs.
573 static void calc_input_response(WMAVoiceContext *s, float *lpcs,
574 int fcb_type, float *coeffs, int remainder)
576 float last_coeff, min = 15.0, max = -15.0;
577 float irange, angle_mul, gain_mul, range, sq;
580 /* Create frequency power spectrum of speech input (i.e. RDFT of LPCs) */
581 s->rdft.rdft_calc(&s->rdft, lpcs);
582 #define log_range(var, assign) do { \
583 float tmp = log10f(assign); var = tmp; \
584 max = FFMAX(max, tmp); min = FFMIN(min, tmp); \
586 log_range(last_coeff, lpcs[1] * lpcs[1]);
587 for (n = 1; n < 64; n++)
588 log_range(lpcs[n], lpcs[n * 2] * lpcs[n * 2] +
589 lpcs[n * 2 + 1] * lpcs[n * 2 + 1]);
590 log_range(lpcs[0], lpcs[0] * lpcs[0]);
593 lpcs[64] = last_coeff;
595 /* Now, use this spectrum to pick out these frequencies with higher
596 * (relative) power/energy (which we then take to be "not noise"),
597 * and set up a table (still in lpc[]) of (relative) gains per frequency.
598 * These frequencies will be maintained, while others ("noise") will be
599 * decreased in the filter output. */
600 irange = 64.0 / range; // so irange*(max-value) is in the range [0, 63]
601 gain_mul = range * (fcb_type == FCB_TYPE_HARDCODED ? (5.0 / 13.0) :
603 angle_mul = gain_mul * (8.0 * M_LN10 / M_PI);
604 for (n = 0; n <= 64; n++) {
607 idx = FFMAX(0, lrint((max - lpcs[n]) * irange) - 1);
608 pwr = wmavoice_denoise_power_table[s->denoise_strength][idx];
609 lpcs[n] = angle_mul * pwr;
611 /* 70.57 =~ 1/log10(1.0331663) */
612 idx = (pwr * gain_mul - 0.0295) * 70.570526123;
613 if (idx > 127) { // fallback if index falls outside table range
614 coeffs[n] = wmavoice_energy_table[127] *
615 powf(1.0331663, idx - 127);
617 coeffs[n] = wmavoice_energy_table[FFMAX(0, idx)];
620 /* calculate the Hilbert transform of the gains, which we do (since this
621 * is a sinus input) by doing a phase shift (in theory, H(sin())=cos()).
622 * Hilbert_Transform(RDFT(x)) = Laplace_Transform(x), which calculates the
623 * "moment" of the LPCs in this filter. */
624 s->dct.dct_calc(&s->dct, lpcs);
625 s->dst.dct_calc(&s->dst, lpcs);
627 /* Split out the coefficient indexes into phase/magnitude pairs */
628 idx = 255 + av_clip(lpcs[64], -255, 255);
629 coeffs[0] = coeffs[0] * s->cos[idx];
630 idx = 255 + av_clip(lpcs[64] - 2 * lpcs[63], -255, 255);
631 last_coeff = coeffs[64] * s->cos[idx];
633 idx = 255 + av_clip(-lpcs[64] - 2 * lpcs[n - 1], -255, 255);
634 coeffs[n * 2 + 1] = coeffs[n] * s->sin[idx];
635 coeffs[n * 2] = coeffs[n] * s->cos[idx];
639 idx = 255 + av_clip( lpcs[64] - 2 * lpcs[n - 1], -255, 255);
640 coeffs[n * 2 + 1] = coeffs[n] * s->sin[idx];
641 coeffs[n * 2] = coeffs[n] * s->cos[idx];
643 coeffs[1] = last_coeff;
645 /* move into real domain */
646 s->irdft.rdft_calc(&s->irdft, coeffs);
648 /* tilt correction and normalize scale */
649 memset(&coeffs[remainder], 0, sizeof(coeffs[0]) * (128 - remainder));
650 if (s->denoise_tilt_corr) {
653 coeffs[remainder - 1] = 0;
654 ff_tilt_compensation(&tilt_mem,
655 -1.8 * tilt_factor(coeffs, remainder - 1),
658 sq = (1.0 / 64.0) * sqrtf(1 / ff_scalarproduct_float_c(coeffs, coeffs, remainder));
659 for (n = 0; n < remainder; n++)
664 * This function applies a Wiener filter on the (noisy) speech signal as
665 * a means to denoise it.
667 * - take RDFT of LPCs to get the power spectrum of the noise + speech;
668 * - using this power spectrum, calculate (for each frequency) the Wiener
669 * filter gain, which depends on the frequency power and desired level
670 * of noise subtraction (when set too high, this leads to artifacts)
671 * We can do this symmetrically over the X-axis (so 0-4kHz is the inverse
673 * - by doing a phase shift, calculate the Hilbert transform of this array
674 * of per-frequency filter-gains to get the filtering coefficients;
675 * - smoothen/normalize/de-tilt these filter coefficients as desired;
676 * - take RDFT of noisy sound, apply the coefficients and take its IRDFT
677 * to get the denoised speech signal;
678 * - the leftover (i.e. output of the IRDFT on denoised speech data beyond
679 * the frame boundary) are saved and applied to subsequent frames by an
680 * overlap-add method (otherwise you get clicking-artifacts).
682 * @param s WMA Voice decoding context
683 * @param fcb_type Frame (codebook) type
684 * @param synth_pf input: the noisy speech signal, output: denoised speech
685 * data; should be 16-byte aligned (for ASM purposes)
686 * @param size size of the speech data
687 * @param lpcs LPCs used to synthesize this frame's speech data
689 static void wiener_denoise(WMAVoiceContext *s, int fcb_type,
690 float *synth_pf, int size,
693 int remainder, lim, n;
695 if (fcb_type != FCB_TYPE_SILENCE) {
696 float *tilted_lpcs = s->tilted_lpcs_pf,
697 *coeffs = s->denoise_coeffs_pf, tilt_mem = 0;
699 tilted_lpcs[0] = 1.0;
700 memcpy(&tilted_lpcs[1], lpcs, sizeof(lpcs[0]) * s->lsps);
701 memset(&tilted_lpcs[s->lsps + 1], 0,
702 sizeof(tilted_lpcs[0]) * (128 - s->lsps - 1));
703 ff_tilt_compensation(&tilt_mem, 0.7 * tilt_factor(lpcs, s->lsps),
704 tilted_lpcs, s->lsps + 2);
706 /* The IRDFT output (127 samples for 7-bit filter) beyond the frame
707 * size is applied to the next frame. All input beyond this is zero,
708 * and thus all output beyond this will go towards zero, hence we can
709 * limit to min(size-1, 127-size) as a performance consideration. */
710 remainder = FFMIN(127 - size, size - 1);
711 calc_input_response(s, tilted_lpcs, fcb_type, coeffs, remainder);
713 /* apply coefficients (in frequency spectrum domain), i.e. complex
714 * number multiplication */
715 memset(&synth_pf[size], 0, sizeof(synth_pf[0]) * (128 - size));
716 s->rdft.rdft_calc(&s->rdft, synth_pf);
717 s->rdft.rdft_calc(&s->rdft, coeffs);
718 synth_pf[0] *= coeffs[0];
719 synth_pf[1] *= coeffs[1];
720 for (n = 1; n < 64; n++) {
721 float v1 = synth_pf[n * 2], v2 = synth_pf[n * 2 + 1];
722 synth_pf[n * 2] = v1 * coeffs[n * 2] - v2 * coeffs[n * 2 + 1];
723 synth_pf[n * 2 + 1] = v2 * coeffs[n * 2] + v1 * coeffs[n * 2 + 1];
725 s->irdft.rdft_calc(&s->irdft, synth_pf);
728 /* merge filter output with the history of previous runs */
729 if (s->denoise_filter_cache_size) {
730 lim = FFMIN(s->denoise_filter_cache_size, size);
731 for (n = 0; n < lim; n++)
732 synth_pf[n] += s->denoise_filter_cache[n];
733 s->denoise_filter_cache_size -= lim;
734 memmove(s->denoise_filter_cache, &s->denoise_filter_cache[size],
735 sizeof(s->denoise_filter_cache[0]) * s->denoise_filter_cache_size);
738 /* move remainder of filter output into a cache for future runs */
739 if (fcb_type != FCB_TYPE_SILENCE) {
740 lim = FFMIN(remainder, s->denoise_filter_cache_size);
741 for (n = 0; n < lim; n++)
742 s->denoise_filter_cache[n] += synth_pf[size + n];
743 if (lim < remainder) {
744 memcpy(&s->denoise_filter_cache[lim], &synth_pf[size + lim],
745 sizeof(s->denoise_filter_cache[0]) * (remainder - lim));
746 s->denoise_filter_cache_size = remainder;
752 * Averaging projection filter, the postfilter used in WMAVoice.
754 * This uses the following steps:
755 * - A zero-synthesis filter (generate excitation from synth signal)
756 * - Kalman smoothing on excitation, based on pitch
757 * - Re-synthesized smoothened output
758 * - Iterative Wiener denoise filter
759 * - Adaptive gain filter
762 * @param s WMAVoice decoding context
763 * @param synth Speech synthesis output (before postfilter)
764 * @param samples Output buffer for filtered samples
765 * @param size Buffer size of synth & samples
766 * @param lpcs Generated LPCs used for speech synthesis
767 * @param zero_exc_pf destination for zero synthesis filter (16-byte aligned)
768 * @param fcb_type Frame type (silence, hardcoded, AW-pulses or FCB-pulses)
769 * @param pitch Pitch of the input signal
771 static void postfilter(WMAVoiceContext *s, const float *synth,
772 float *samples, int size,
773 const float *lpcs, float *zero_exc_pf,
774 int fcb_type, int pitch)
776 float synth_filter_in_buf[MAX_FRAMESIZE / 2],
777 *synth_pf = &s->synth_filter_out_buf[MAX_LSPS_ALIGN16],
778 *synth_filter_in = zero_exc_pf;
780 av_assert0(size <= MAX_FRAMESIZE / 2);
782 /* generate excitation from input signal */
783 ff_celp_lp_zero_synthesis_filterf(zero_exc_pf, lpcs, synth, size, s->lsps);
785 if (fcb_type >= FCB_TYPE_AW_PULSES &&
786 !kalman_smoothen(s, pitch, zero_exc_pf, synth_filter_in_buf, size))
787 synth_filter_in = synth_filter_in_buf;
789 /* re-synthesize speech after smoothening, and keep history */
790 ff_celp_lp_synthesis_filterf(synth_pf, lpcs,
791 synth_filter_in, size, s->lsps);
792 memcpy(&synth_pf[-s->lsps], &synth_pf[size - s->lsps],
793 sizeof(synth_pf[0]) * s->lsps);
795 wiener_denoise(s, fcb_type, synth_pf, size, lpcs);
797 adaptive_gain_control(samples, synth_pf, synth, size, 0.99,
800 if (s->dc_level > 8) {
801 /* remove ultra-low frequency DC noise / highpass filter;
802 * coefficients are identical to those used in SIPR decoding,
803 * and very closely resemble those used in AMR-NB decoding. */
804 ff_acelp_apply_order_2_transfer_function(samples, samples,
805 (const float[2]) { -1.99997, 1.0 },
806 (const float[2]) { -1.9330735188, 0.93589198496 },
807 0.93980580475, s->dcf_mem, size);
816 * @param lsps output pointer to the array that will hold the LSPs
817 * @param num number of LSPs to be dequantized
818 * @param values quantized values, contains n_stages values
819 * @param sizes range (i.e. max value) of each quantized value
820 * @param n_stages number of dequantization runs
821 * @param table dequantization table to be used
822 * @param mul_q LSF multiplier
823 * @param base_q base (lowest) LSF values
825 static void dequant_lsps(double *lsps, int num,
826 const uint16_t *values,
827 const uint16_t *sizes,
828 int n_stages, const uint8_t *table,
830 const double *base_q)
834 memset(lsps, 0, num * sizeof(*lsps));
835 for (n = 0; n < n_stages; n++) {
836 const uint8_t *t_off = &table[values[n] * num];
837 double base = base_q[n], mul = mul_q[n];
839 for (m = 0; m < num; m++)
840 lsps[m] += base + mul * t_off[m];
842 table += sizes[n] * num;
847 * @name LSP dequantization routines
848 * LSP dequantization routines, for 10/16LSPs and independent/residual coding.
849 * @note we assume enough bits are available, caller should check.
850 * lsp10i() consumes 24 bits; lsp10r() consumes an additional 24 bits;
851 * lsp16i() consumes 34 bits; lsp16r() consumes an additional 26 bits.
855 * Parse 10 independently-coded LSPs.
857 static void dequant_lsp10i(GetBitContext *gb, double *lsps)
859 static const uint16_t vec_sizes[4] = { 256, 64, 32, 32 };
860 static const double mul_lsf[4] = {
861 5.2187144800e-3, 1.4626986422e-3,
862 9.6179549166e-4, 1.1325736225e-3
864 static const double base_lsf[4] = {
865 M_PI * -2.15522e-1, M_PI * -6.1646e-2,
866 M_PI * -3.3486e-2, M_PI * -5.7408e-2
870 v[0] = get_bits(gb, 8);
871 v[1] = get_bits(gb, 6);
872 v[2] = get_bits(gb, 5);
873 v[3] = get_bits(gb, 5);
875 dequant_lsps(lsps, 10, v, vec_sizes, 4, wmavoice_dq_lsp10i,
880 * Parse 10 independently-coded LSPs, and then derive the tables to
881 * generate LSPs for the other frames from them (residual coding).
883 static void dequant_lsp10r(GetBitContext *gb,
884 double *i_lsps, const double *old,
885 double *a1, double *a2, int q_mode)
887 static const uint16_t vec_sizes[3] = { 128, 64, 64 };
888 static const double mul_lsf[3] = {
889 2.5807601174e-3, 1.2354460219e-3, 1.1763821673e-3
891 static const double base_lsf[3] = {
892 M_PI * -1.07448e-1, M_PI * -5.2706e-2, M_PI * -5.1634e-2
894 const float (*ipol_tab)[2][10] = q_mode ?
895 wmavoice_lsp10_intercoeff_b : wmavoice_lsp10_intercoeff_a;
896 uint16_t interpol, v[3];
899 dequant_lsp10i(gb, i_lsps);
901 interpol = get_bits(gb, 5);
902 v[0] = get_bits(gb, 7);
903 v[1] = get_bits(gb, 6);
904 v[2] = get_bits(gb, 6);
906 for (n = 0; n < 10; n++) {
907 double delta = old[n] - i_lsps[n];
908 a1[n] = ipol_tab[interpol][0][n] * delta + i_lsps[n];
909 a1[10 + n] = ipol_tab[interpol][1][n] * delta + i_lsps[n];
912 dequant_lsps(a2, 20, v, vec_sizes, 3, wmavoice_dq_lsp10r,
917 * Parse 16 independently-coded LSPs.
919 static void dequant_lsp16i(GetBitContext *gb, double *lsps)
921 static const uint16_t vec_sizes[5] = { 256, 64, 128, 64, 128 };
922 static const double mul_lsf[5] = {
923 3.3439586280e-3, 6.9908173703e-4,
924 3.3216608306e-3, 1.0334960326e-3,
927 static const double base_lsf[5] = {
928 M_PI * -1.27576e-1, M_PI * -2.4292e-2,
929 M_PI * -1.28094e-1, M_PI * -3.2128e-2,
934 v[0] = get_bits(gb, 8);
935 v[1] = get_bits(gb, 6);
936 v[2] = get_bits(gb, 7);
937 v[3] = get_bits(gb, 6);
938 v[4] = get_bits(gb, 7);
940 dequant_lsps( lsps, 5, v, vec_sizes, 2,
941 wmavoice_dq_lsp16i1, mul_lsf, base_lsf);
942 dequant_lsps(&lsps[5], 5, &v[2], &vec_sizes[2], 2,
943 wmavoice_dq_lsp16i2, &mul_lsf[2], &base_lsf[2]);
944 dequant_lsps(&lsps[10], 6, &v[4], &vec_sizes[4], 1,
945 wmavoice_dq_lsp16i3, &mul_lsf[4], &base_lsf[4]);
949 * Parse 16 independently-coded LSPs, and then derive the tables to
950 * generate LSPs for the other frames from them (residual coding).
952 static void dequant_lsp16r(GetBitContext *gb,
953 double *i_lsps, const double *old,
954 double *a1, double *a2, int q_mode)
956 static const uint16_t vec_sizes[3] = { 128, 128, 128 };
957 static const double mul_lsf[3] = {
958 1.2232979501e-3, 1.4062241527e-3, 1.6114744851e-3
960 static const double base_lsf[3] = {
961 M_PI * -5.5830e-2, M_PI * -5.2908e-2, M_PI * -5.4776e-2
963 const float (*ipol_tab)[2][16] = q_mode ?
964 wmavoice_lsp16_intercoeff_b : wmavoice_lsp16_intercoeff_a;
965 uint16_t interpol, v[3];
968 dequant_lsp16i(gb, i_lsps);
970 interpol = get_bits(gb, 5);
971 v[0] = get_bits(gb, 7);
972 v[1] = get_bits(gb, 7);
973 v[2] = get_bits(gb, 7);
975 for (n = 0; n < 16; n++) {
976 double delta = old[n] - i_lsps[n];
977 a1[n] = ipol_tab[interpol][0][n] * delta + i_lsps[n];
978 a1[16 + n] = ipol_tab[interpol][1][n] * delta + i_lsps[n];
981 dequant_lsps( a2, 10, v, vec_sizes, 1,
982 wmavoice_dq_lsp16r1, mul_lsf, base_lsf);
983 dequant_lsps(&a2[10], 10, &v[1], &vec_sizes[1], 1,
984 wmavoice_dq_lsp16r2, &mul_lsf[1], &base_lsf[1]);
985 dequant_lsps(&a2[20], 12, &v[2], &vec_sizes[2], 1,
986 wmavoice_dq_lsp16r3, &mul_lsf[2], &base_lsf[2]);
991 * @name Pitch-adaptive window coding functions
992 * The next few functions are for pitch-adaptive window coding.
996 * Parse the offset of the first pitch-adaptive window pulses, and
997 * the distribution of pulses between the two blocks in this frame.
998 * @param s WMA Voice decoding context private data
999 * @param gb bit I/O context
1000 * @param pitch pitch for each block in this frame
1002 static void aw_parse_coords(WMAVoiceContext *s, GetBitContext *gb,
1005 static const int16_t start_offset[94] = {
1006 -11, -9, -7, -5, -3, -1, 1, 3, 5, 7, 9, 11,
1007 13, 15, 18, 17, 19, 20, 21, 22, 23, 24, 25, 26,
1008 27, 28, 29, 30, 31, 32, 33, 35, 37, 39, 41, 43,
1009 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65, 67,
1010 69, 71, 73, 75, 77, 79, 81, 83, 85, 87, 89, 91,
1011 93, 95, 97, 99, 101, 103, 105, 107, 109, 111, 113, 115,
1012 117, 119, 121, 123, 125, 127, 129, 131, 133, 135, 137, 139,
1013 141, 143, 145, 147, 149, 151, 153, 155, 157, 159
1017 /* position of pulse */
1018 s->aw_idx_is_ext = 0;
1019 if ((bits = get_bits(gb, 6)) >= 54) {
1020 s->aw_idx_is_ext = 1;
1021 bits += (bits - 54) * 3 + get_bits(gb, 2);
1024 /* for a repeated pulse at pulse_off with a pitch_lag of pitch[], count
1025 * the distribution of the pulses in each block contained in this frame. */
1026 s->aw_pulse_range = FFMIN(pitch[0], pitch[1]) > 32 ? 24 : 16;
1027 for (offset = start_offset[bits]; offset < 0; offset += pitch[0]) ;
1028 s->aw_n_pulses[0] = (pitch[0] - 1 + MAX_FRAMESIZE / 2 - offset) / pitch[0];
1029 s->aw_first_pulse_off[0] = offset - s->aw_pulse_range / 2;
1030 offset += s->aw_n_pulses[0] * pitch[0];
1031 s->aw_n_pulses[1] = (pitch[1] - 1 + MAX_FRAMESIZE - offset) / pitch[1];
1032 s->aw_first_pulse_off[1] = offset - (MAX_FRAMESIZE + s->aw_pulse_range) / 2;
1034 /* if continuing from a position before the block, reset position to
1035 * start of block (when corrected for the range over which it can be
1036 * spread in aw_pulse_set1()). */
1037 if (start_offset[bits] < MAX_FRAMESIZE / 2) {
1038 while (s->aw_first_pulse_off[1] - pitch[1] + s->aw_pulse_range > 0)
1039 s->aw_first_pulse_off[1] -= pitch[1];
1040 if (start_offset[bits] < 0)
1041 while (s->aw_first_pulse_off[0] - pitch[0] + s->aw_pulse_range > 0)
1042 s->aw_first_pulse_off[0] -= pitch[0];
1047 * Apply second set of pitch-adaptive window pulses.
1048 * @param s WMA Voice decoding context private data
1049 * @param gb bit I/O context
1050 * @param block_idx block index in frame [0, 1]
1051 * @param fcb structure containing fixed codebook vector info
1053 static void aw_pulse_set2(WMAVoiceContext *s, GetBitContext *gb,
1054 int block_idx, AMRFixed *fcb)
1056 uint16_t use_mask_mem[9]; // only 5 are used, rest is padding
1057 uint16_t *use_mask = use_mask_mem + 2;
1058 /* in this function, idx is the index in the 80-bit (+ padding) use_mask
1059 * bit-array. Since use_mask consists of 16-bit values, the lower 4 bits
1060 * of idx are the position of the bit within a particular item in the
1061 * array (0 being the most significant bit, and 15 being the least
1062 * significant bit), and the remainder (>> 4) is the index in the
1063 * use_mask[]-array. This is faster and uses less memory than using a
1064 * 80-byte/80-int array. */
1065 int pulse_off = s->aw_first_pulse_off[block_idx],
1066 pulse_start, n, idx, range, aidx, start_off = 0;
1068 /* set offset of first pulse to within this block */
1069 if (s->aw_n_pulses[block_idx] > 0)
1070 while (pulse_off + s->aw_pulse_range < 1)
1071 pulse_off += fcb->pitch_lag;
1073 /* find range per pulse */
1074 if (s->aw_n_pulses[0] > 0) {
1075 if (block_idx == 0) {
1077 } else /* block_idx = 1 */ {
1079 if (s->aw_n_pulses[block_idx] > 0)
1080 pulse_off = s->aw_next_pulse_off_cache;
1084 pulse_start = s->aw_n_pulses[block_idx] > 0 ? pulse_off - range / 2 : 0;
1086 /* aw_pulse_set1() already applies pulses around pulse_off (to be exactly,
1087 * in the range of [pulse_off, pulse_off + s->aw_pulse_range], and thus
1088 * we exclude that range from being pulsed again in this function. */
1089 memset(&use_mask[-2], 0, 2 * sizeof(use_mask[0]));
1090 memset( use_mask, -1, 5 * sizeof(use_mask[0]));
1091 memset(&use_mask[5], 0, 2 * sizeof(use_mask[0]));
1092 if (s->aw_n_pulses[block_idx] > 0)
1093 for (idx = pulse_off; idx < MAX_FRAMESIZE / 2; idx += fcb->pitch_lag) {
1094 int excl_range = s->aw_pulse_range; // always 16 or 24
1095 uint16_t *use_mask_ptr = &use_mask[idx >> 4];
1096 int first_sh = 16 - (idx & 15);
1097 *use_mask_ptr++ &= 0xFFFFu << first_sh;
1098 excl_range -= first_sh;
1099 if (excl_range >= 16) {
1100 *use_mask_ptr++ = 0;
1101 *use_mask_ptr &= 0xFFFF >> (excl_range - 16);
1103 *use_mask_ptr &= 0xFFFF >> excl_range;
1106 /* find the 'aidx'th offset that is not excluded */
1107 aidx = get_bits(gb, s->aw_n_pulses[0] > 0 ? 5 - 2 * block_idx : 4);
1108 for (n = 0; n <= aidx; pulse_start++) {
1109 for (idx = pulse_start; idx < 0; idx += fcb->pitch_lag) ;
1110 if (idx >= MAX_FRAMESIZE / 2) { // find from zero
1111 if (use_mask[0]) idx = 0x0F;
1112 else if (use_mask[1]) idx = 0x1F;
1113 else if (use_mask[2]) idx = 0x2F;
1114 else if (use_mask[3]) idx = 0x3F;
1115 else if (use_mask[4]) idx = 0x4F;
1117 idx -= av_log2_16bit(use_mask[idx >> 4]);
1119 if (use_mask[idx >> 4] & (0x8000 >> (idx & 15))) {
1120 use_mask[idx >> 4] &= ~(0x8000 >> (idx & 15));
1126 fcb->x[fcb->n] = start_off;
1127 fcb->y[fcb->n] = get_bits1(gb) ? -1.0 : 1.0;
1130 /* set offset for next block, relative to start of that block */
1131 n = (MAX_FRAMESIZE / 2 - start_off) % fcb->pitch_lag;
1132 s->aw_next_pulse_off_cache = n ? fcb->pitch_lag - n : 0;
1136 * Apply first set of pitch-adaptive window pulses.
1137 * @param s WMA Voice decoding context private data
1138 * @param gb bit I/O context
1139 * @param block_idx block index in frame [0, 1]
1140 * @param fcb storage location for fixed codebook pulse info
1142 static void aw_pulse_set1(WMAVoiceContext *s, GetBitContext *gb,
1143 int block_idx, AMRFixed *fcb)
1145 int val = get_bits(gb, 12 - 2 * (s->aw_idx_is_ext && !block_idx));
1148 if (s->aw_n_pulses[block_idx] > 0) {
1149 int n, v_mask, i_mask, sh, n_pulses;
1151 if (s->aw_pulse_range == 24) { // 3 pulses, 1:sign + 3:index each
1156 } else { // 4 pulses, 1:sign + 2:index each
1163 for (n = n_pulses - 1; n >= 0; n--, val >>= sh) {
1164 fcb->y[fcb->n] = (val & v_mask) ? -1.0 : 1.0;
1165 fcb->x[fcb->n] = (val & i_mask) * n_pulses + n +
1166 s->aw_first_pulse_off[block_idx];
1167 while (fcb->x[fcb->n] < 0)
1168 fcb->x[fcb->n] += fcb->pitch_lag;
1169 if (fcb->x[fcb->n] < MAX_FRAMESIZE / 2)
1173 int num2 = (val & 0x1FF) >> 1, delta, idx;
1175 if (num2 < 1 * 79) { delta = 1; idx = num2 + 1; }
1176 else if (num2 < 2 * 78) { delta = 3; idx = num2 + 1 - 1 * 77; }
1177 else if (num2 < 3 * 77) { delta = 5; idx = num2 + 1 - 2 * 76; }
1178 else { delta = 7; idx = num2 + 1 - 3 * 75; }
1179 v = (val & 0x200) ? -1.0 : 1.0;
1181 fcb->no_repeat_mask |= 3 << fcb->n;
1182 fcb->x[fcb->n] = idx - delta;
1184 fcb->x[fcb->n + 1] = idx;
1185 fcb->y[fcb->n + 1] = (val & 1) ? -v : v;
1193 * Generate a random number from frame_cntr and block_idx, which will lief
1194 * in the range [0, 1000 - block_size] (so it can be used as an index in a
1195 * table of size 1000 of which you want to read block_size entries).
1197 * @param frame_cntr current frame number
1198 * @param block_num current block index
1199 * @param block_size amount of entries we want to read from a table
1200 * that has 1000 entries
1201 * @return a (non-)random number in the [0, 1000 - block_size] range.
1203 static int pRNG(int frame_cntr, int block_num, int block_size)
1205 /* array to simplify the calculation of z:
1206 * y = (x % 9) * 5 + 6;
1207 * z = (49995 * x) / y;
1208 * Since y only has 9 values, we can remove the division by using a
1209 * LUT and using FASTDIV-style divisions. For each of the 9 values
1210 * of y, we can rewrite z as:
1211 * z = x * (49995 / y) + x * ((49995 % y) / y)
1212 * In this table, each col represents one possible value of y, the
1213 * first number is 49995 / y, and the second is the FASTDIV variant
1214 * of 49995 % y / y. */
1215 static const unsigned int div_tbl[9][2] = {
1216 { 8332, 3 * 715827883U }, // y = 6
1217 { 4545, 0 * 390451573U }, // y = 11
1218 { 3124, 11 * 268435456U }, // y = 16
1219 { 2380, 15 * 204522253U }, // y = 21
1220 { 1922, 23 * 165191050U }, // y = 26
1221 { 1612, 23 * 138547333U }, // y = 31
1222 { 1388, 27 * 119304648U }, // y = 36
1223 { 1219, 16 * 104755300U }, // y = 41
1224 { 1086, 39 * 93368855U } // y = 46
1226 unsigned int z, y, x = MUL16(block_num, 1877) + frame_cntr;
1227 if (x >= 0xFFFF) x -= 0xFFFF; // max value of x is 8*1877+0xFFFE=0x13AA6,
1228 // so this is effectively a modulo (%)
1229 y = x - 9 * MULH(477218589, x); // x % 9
1230 z = (uint16_t) (x * div_tbl[y][0] + UMULH(x, div_tbl[y][1]));
1231 // z = x * 49995 / (y * 5 + 6)
1232 return z % (1000 - block_size);
1236 * Parse hardcoded signal for a single block.
1237 * @note see #synth_block().
1239 static void synth_block_hardcoded(WMAVoiceContext *s, GetBitContext *gb,
1240 int block_idx, int size,
1241 const struct frame_type_desc *frame_desc,
1247 av_assert0(size <= MAX_FRAMESIZE);
1249 /* Set the offset from which we start reading wmavoice_std_codebook */
1250 if (frame_desc->fcb_type == FCB_TYPE_SILENCE) {
1251 r_idx = pRNG(s->frame_cntr, block_idx, size);
1252 gain = s->silence_gain;
1253 } else /* FCB_TYPE_HARDCODED */ {
1254 r_idx = get_bits(gb, 8);
1255 gain = wmavoice_gain_universal[get_bits(gb, 6)];
1258 /* Clear gain prediction parameters */
1259 memset(s->gain_pred_err, 0, sizeof(s->gain_pred_err));
1261 /* Apply gain to hardcoded codebook and use that as excitation signal */
1262 for (n = 0; n < size; n++)
1263 excitation[n] = wmavoice_std_codebook[r_idx + n] * gain;
1267 * Parse FCB/ACB signal for a single block.
1268 * @note see #synth_block().
1270 static void synth_block_fcb_acb(WMAVoiceContext *s, GetBitContext *gb,
1271 int block_idx, int size,
1272 int block_pitch_sh2,
1273 const struct frame_type_desc *frame_desc,
1276 static const float gain_coeff[6] = {
1277 0.8169, -0.06545, 0.1726, 0.0185, -0.0359, 0.0458
1279 float pulses[MAX_FRAMESIZE / 2], pred_err, acb_gain, fcb_gain;
1280 int n, idx, gain_weight;
1283 av_assert0(size <= MAX_FRAMESIZE / 2);
1284 memset(pulses, 0, sizeof(*pulses) * size);
1286 fcb.pitch_lag = block_pitch_sh2 >> 2;
1287 fcb.pitch_fac = 1.0;
1288 fcb.no_repeat_mask = 0;
1291 /* For the other frame types, this is where we apply the innovation
1292 * (fixed) codebook pulses of the speech signal. */
1293 if (frame_desc->fcb_type == FCB_TYPE_AW_PULSES) {
1294 aw_pulse_set1(s, gb, block_idx, &fcb);
1295 aw_pulse_set2(s, gb, block_idx, &fcb);
1296 } else /* FCB_TYPE_EXC_PULSES */ {
1297 int offset_nbits = 5 - frame_desc->log_n_blocks;
1299 fcb.no_repeat_mask = -1;
1300 /* similar to ff_decode_10_pulses_35bits(), but with single pulses
1301 * (instead of double) for a subset of pulses */
1302 for (n = 0; n < 5; n++) {
1306 sign = get_bits1(gb) ? 1.0 : -1.0;
1307 pos1 = get_bits(gb, offset_nbits);
1308 fcb.x[fcb.n] = n + 5 * pos1;
1309 fcb.y[fcb.n++] = sign;
1310 if (n < frame_desc->dbl_pulses) {
1311 pos2 = get_bits(gb, offset_nbits);
1312 fcb.x[fcb.n] = n + 5 * pos2;
1313 fcb.y[fcb.n++] = (pos1 < pos2) ? -sign : sign;
1317 ff_set_fixed_vector(pulses, &fcb, 1.0, size);
1319 /* Calculate gain for adaptive & fixed codebook signal.
1320 * see ff_amr_set_fixed_gain(). */
1321 idx = get_bits(gb, 7);
1322 fcb_gain = expf(ff_scalarproduct_float_c(s->gain_pred_err, gain_coeff, 6) -
1323 5.2409161640 + wmavoice_gain_codebook_fcb[idx]);
1324 acb_gain = wmavoice_gain_codebook_acb[idx];
1325 pred_err = av_clipf(wmavoice_gain_codebook_fcb[idx],
1326 -2.9957322736 /* log(0.05) */,
1327 1.6094379124 /* log(5.0) */);
1329 gain_weight = 8 >> frame_desc->log_n_blocks;
1330 memmove(&s->gain_pred_err[gain_weight], s->gain_pred_err,
1331 sizeof(*s->gain_pred_err) * (6 - gain_weight));
1332 for (n = 0; n < gain_weight; n++)
1333 s->gain_pred_err[n] = pred_err;
1335 /* Calculation of adaptive codebook */
1336 if (frame_desc->acb_type == ACB_TYPE_ASYMMETRIC) {
1338 for (n = 0; n < size; n += len) {
1340 int abs_idx = block_idx * size + n;
1341 int pitch_sh16 = (s->last_pitch_val << 16) +
1342 s->pitch_diff_sh16 * abs_idx;
1343 int pitch = (pitch_sh16 + 0x6FFF) >> 16;
1344 int idx_sh16 = ((pitch << 16) - pitch_sh16) * 8 + 0x58000;
1345 idx = idx_sh16 >> 16;
1346 if (s->pitch_diff_sh16) {
1347 if (s->pitch_diff_sh16 > 0) {
1348 next_idx_sh16 = (idx_sh16) &~ 0xFFFF;
1350 next_idx_sh16 = (idx_sh16 + 0x10000) &~ 0xFFFF;
1351 len = av_clip((idx_sh16 - next_idx_sh16) / s->pitch_diff_sh16 / 8,
1356 ff_acelp_interpolatef(&excitation[n], &excitation[n - pitch],
1357 wmavoice_ipol1_coeffs, 17,
1360 } else /* ACB_TYPE_HAMMING */ {
1361 int block_pitch = block_pitch_sh2 >> 2;
1362 idx = block_pitch_sh2 & 3;
1364 ff_acelp_interpolatef(excitation, &excitation[-block_pitch],
1365 wmavoice_ipol2_coeffs, 4,
1368 av_memcpy_backptr((uint8_t *) excitation, sizeof(float) * block_pitch,
1369 sizeof(float) * size);
1372 /* Interpolate ACB/FCB and use as excitation signal */
1373 ff_weighted_vector_sumf(excitation, excitation, pulses,
1374 acb_gain, fcb_gain, size);
1378 * Parse data in a single block.
1379 * @note we assume enough bits are available, caller should check.
1381 * @param s WMA Voice decoding context private data
1382 * @param gb bit I/O context
1383 * @param block_idx index of the to-be-read block
1384 * @param size amount of samples to be read in this block
1385 * @param block_pitch_sh2 pitch for this block << 2
1386 * @param lsps LSPs for (the end of) this frame
1387 * @param prev_lsps LSPs for the last frame
1388 * @param frame_desc frame type descriptor
1389 * @param excitation target memory for the ACB+FCB interpolated signal
1390 * @param synth target memory for the speech synthesis filter output
1391 * @return 0 on success, <0 on error.
1393 static void synth_block(WMAVoiceContext *s, GetBitContext *gb,
1394 int block_idx, int size,
1395 int block_pitch_sh2,
1396 const double *lsps, const double *prev_lsps,
1397 const struct frame_type_desc *frame_desc,
1398 float *excitation, float *synth)
1400 double i_lsps[MAX_LSPS];
1401 float lpcs[MAX_LSPS];
1405 if (frame_desc->acb_type == ACB_TYPE_NONE)
1406 synth_block_hardcoded(s, gb, block_idx, size, frame_desc, excitation);
1408 synth_block_fcb_acb(s, gb, block_idx, size, block_pitch_sh2,
1409 frame_desc, excitation);
1411 /* convert interpolated LSPs to LPCs */
1412 fac = (block_idx + 0.5) / frame_desc->n_blocks;
1413 for (n = 0; n < s->lsps; n++) // LSF -> LSP
1414 i_lsps[n] = cos(prev_lsps[n] + fac * (lsps[n] - prev_lsps[n]));
1415 ff_acelp_lspd2lpc(i_lsps, lpcs, s->lsps >> 1);
1417 /* Speech synthesis */
1418 ff_celp_lp_synthesis_filterf(synth, lpcs, excitation, size, s->lsps);
1422 * Synthesize output samples for a single frame.
1423 * @note we assume enough bits are available, caller should check.
1425 * @param ctx WMA Voice decoder context
1426 * @param gb bit I/O context (s->gb or one for cross-packet superframes)
1427 * @param frame_idx Frame number within superframe [0-2]
1428 * @param samples pointer to output sample buffer, has space for at least 160
1430 * @param lsps LSP array
1431 * @param prev_lsps array of previous frame's LSPs
1432 * @param excitation target buffer for excitation signal
1433 * @param synth target buffer for synthesized speech data
1434 * @return 0 on success, <0 on error.
1436 static int synth_frame(AVCodecContext *ctx, GetBitContext *gb, int frame_idx,
1438 const double *lsps, const double *prev_lsps,
1439 float *excitation, float *synth)
1441 WMAVoiceContext *s = ctx->priv_data;
1442 int n, n_blocks_x2, log_n_blocks_x2, cur_pitch_val;
1443 int pitch[MAX_BLOCKS], last_block_pitch;
1445 /* Parse frame type ("frame header"), see frame_descs */
1446 int bd_idx = s->vbm_tree[get_vlc2(gb, frame_type_vlc.table, 6, 3)], block_nsamples;
1449 av_log(ctx, AV_LOG_ERROR,
1450 "Invalid frame type VLC code, skipping\n");
1454 block_nsamples = MAX_FRAMESIZE / frame_descs[bd_idx].n_blocks;
1456 /* Pitch calculation for ACB_TYPE_ASYMMETRIC ("pitch-per-frame") */
1457 if (frame_descs[bd_idx].acb_type == ACB_TYPE_ASYMMETRIC) {
1458 /* Pitch is provided per frame, which is interpreted as the pitch of
1459 * the last sample of the last block of this frame. We can interpolate
1460 * the pitch of other blocks (and even pitch-per-sample) by gradually
1461 * incrementing/decrementing prev_frame_pitch to cur_pitch_val. */
1462 n_blocks_x2 = frame_descs[bd_idx].n_blocks << 1;
1463 log_n_blocks_x2 = frame_descs[bd_idx].log_n_blocks + 1;
1464 cur_pitch_val = s->min_pitch_val + get_bits(gb, s->pitch_nbits);
1465 cur_pitch_val = FFMIN(cur_pitch_val, s->max_pitch_val - 1);
1466 if (s->last_acb_type == ACB_TYPE_NONE ||
1467 20 * abs(cur_pitch_val - s->last_pitch_val) >
1468 (cur_pitch_val + s->last_pitch_val))
1469 s->last_pitch_val = cur_pitch_val;
1471 /* pitch per block */
1472 for (n = 0; n < frame_descs[bd_idx].n_blocks; n++) {
1473 int fac = n * 2 + 1;
1475 pitch[n] = (MUL16(fac, cur_pitch_val) +
1476 MUL16((n_blocks_x2 - fac), s->last_pitch_val) +
1477 frame_descs[bd_idx].n_blocks) >> log_n_blocks_x2;
1480 /* "pitch-diff-per-sample" for calculation of pitch per sample */
1481 s->pitch_diff_sh16 =
1482 ((cur_pitch_val - s->last_pitch_val) << 16) / MAX_FRAMESIZE;
1485 /* Global gain (if silence) and pitch-adaptive window coordinates */
1486 switch (frame_descs[bd_idx].fcb_type) {
1487 case FCB_TYPE_SILENCE:
1488 s->silence_gain = wmavoice_gain_silence[get_bits(gb, 8)];
1490 case FCB_TYPE_AW_PULSES:
1491 aw_parse_coords(s, gb, pitch);
1495 for (n = 0; n < frame_descs[bd_idx].n_blocks; n++) {
1498 /* Pitch calculation for ACB_TYPE_HAMMING ("pitch-per-block") */
1499 switch (frame_descs[bd_idx].acb_type) {
1500 case ACB_TYPE_HAMMING: {
1501 /* Pitch is given per block. Per-block pitches are encoded as an
1502 * absolute value for the first block, and then delta values
1503 * relative to this value) for all subsequent blocks. The scale of
1504 * this pitch value is semi-logaritmic compared to its use in the
1505 * decoder, so we convert it to normal scale also. */
1507 t1 = (s->block_conv_table[1] - s->block_conv_table[0]) << 2,
1508 t2 = (s->block_conv_table[2] - s->block_conv_table[1]) << 1,
1509 t3 = s->block_conv_table[3] - s->block_conv_table[2] + 1;
1512 block_pitch = get_bits(gb, s->block_pitch_nbits);
1514 block_pitch = last_block_pitch - s->block_delta_pitch_hrange +
1515 get_bits(gb, s->block_delta_pitch_nbits);
1516 /* Convert last_ so that any next delta is within _range */
1517 last_block_pitch = av_clip(block_pitch,
1518 s->block_delta_pitch_hrange,
1519 s->block_pitch_range -
1520 s->block_delta_pitch_hrange);
1522 /* Convert semi-log-style scale back to normal scale */
1523 if (block_pitch < t1) {
1524 bl_pitch_sh2 = (s->block_conv_table[0] << 2) + block_pitch;
1527 if (block_pitch < t2) {
1529 (s->block_conv_table[1] << 2) + (block_pitch << 1);
1532 if (block_pitch < t3) {
1534 (s->block_conv_table[2] + block_pitch) << 2;
1536 bl_pitch_sh2 = s->block_conv_table[3] << 2;
1539 pitch[n] = bl_pitch_sh2 >> 2;
1543 case ACB_TYPE_ASYMMETRIC: {
1544 bl_pitch_sh2 = pitch[n] << 2;
1548 default: // ACB_TYPE_NONE has no pitch
1553 synth_block(s, gb, n, block_nsamples, bl_pitch_sh2,
1554 lsps, prev_lsps, &frame_descs[bd_idx],
1555 &excitation[n * block_nsamples],
1556 &synth[n * block_nsamples]);
1559 /* Averaging projection filter, if applicable. Else, just copy samples
1560 * from synthesis buffer */
1562 double i_lsps[MAX_LSPS];
1563 float lpcs[MAX_LSPS];
1565 for (n = 0; n < s->lsps; n++) // LSF -> LSP
1566 i_lsps[n] = cos(0.5 * (prev_lsps[n] + lsps[n]));
1567 ff_acelp_lspd2lpc(i_lsps, lpcs, s->lsps >> 1);
1568 postfilter(s, synth, samples, 80, lpcs,
1569 &s->zero_exc_pf[s->history_nsamples + MAX_FRAMESIZE * frame_idx],
1570 frame_descs[bd_idx].fcb_type, pitch[0]);
1572 for (n = 0; n < s->lsps; n++) // LSF -> LSP
1573 i_lsps[n] = cos(lsps[n]);
1574 ff_acelp_lspd2lpc(i_lsps, lpcs, s->lsps >> 1);
1575 postfilter(s, &synth[80], &samples[80], 80, lpcs,
1576 &s->zero_exc_pf[s->history_nsamples + MAX_FRAMESIZE * frame_idx + 80],
1577 frame_descs[bd_idx].fcb_type, pitch[0]);
1579 memcpy(samples, synth, 160 * sizeof(synth[0]));
1581 /* Cache values for next frame */
1583 if (s->frame_cntr >= 0xFFFF) s->frame_cntr -= 0xFFFF; // i.e. modulo (%)
1584 s->last_acb_type = frame_descs[bd_idx].acb_type;
1585 switch (frame_descs[bd_idx].acb_type) {
1587 s->last_pitch_val = 0;
1589 case ACB_TYPE_ASYMMETRIC:
1590 s->last_pitch_val = cur_pitch_val;
1592 case ACB_TYPE_HAMMING:
1593 s->last_pitch_val = pitch[frame_descs[bd_idx].n_blocks - 1];
1601 * Ensure minimum value for first item, maximum value for last value,
1602 * proper spacing between each value and proper ordering.
1604 * @param lsps array of LSPs
1605 * @param num size of LSP array
1607 * @note basically a double version of #ff_acelp_reorder_lsf(), might be
1608 * useful to put in a generic location later on. Parts are also
1609 * present in #ff_set_min_dist_lsf() + #ff_sort_nearly_sorted_floats(),
1610 * which is in float.
1612 static void stabilize_lsps(double *lsps, int num)
1616 /* set minimum value for first, maximum value for last and minimum
1617 * spacing between LSF values.
1618 * Very similar to ff_set_min_dist_lsf(), but in double. */
1619 lsps[0] = FFMAX(lsps[0], 0.0015 * M_PI);
1620 for (n = 1; n < num; n++)
1621 lsps[n] = FFMAX(lsps[n], lsps[n - 1] + 0.0125 * M_PI);
1622 lsps[num - 1] = FFMIN(lsps[num - 1], 0.9985 * M_PI);
1624 /* reorder (looks like one-time / non-recursed bubblesort).
1625 * Very similar to ff_sort_nearly_sorted_floats(), but in double. */
1626 for (n = 1; n < num; n++) {
1627 if (lsps[n] < lsps[n - 1]) {
1628 for (m = 1; m < num; m++) {
1629 double tmp = lsps[m];
1630 for (l = m - 1; l >= 0; l--) {
1631 if (lsps[l] <= tmp) break;
1632 lsps[l + 1] = lsps[l];
1642 * Test if there's enough bits to read 1 superframe.
1644 * @param orig_gb bit I/O context used for reading. This function
1645 * does not modify the state of the bitreader; it
1646 * only uses it to copy the current stream position
1647 * @param s WMA Voice decoding context private data
1648 * @return -1 if unsupported, 1 on not enough bits or 0 if OK.
1650 static int check_bits_for_superframe(GetBitContext *orig_gb,
1653 GetBitContext s_gb, *gb = &s_gb;
1654 int n, need_bits, bd_idx;
1655 const struct frame_type_desc *frame_desc;
1657 /* initialize a copy */
1658 init_get_bits(gb, orig_gb->buffer, orig_gb->size_in_bits);
1659 skip_bits_long(gb, get_bits_count(orig_gb));
1660 av_assert1(get_bits_left(gb) == get_bits_left(orig_gb));
1662 /* superframe header */
1663 if (get_bits_left(gb) < 14)
1666 return -1; // WMAPro-in-WMAVoice superframe
1667 if (get_bits1(gb)) skip_bits(gb, 12); // number of samples in superframe
1668 if (s->has_residual_lsps) { // residual LSPs (for all frames)
1669 if (get_bits_left(gb) < s->sframe_lsp_bitsize)
1671 skip_bits_long(gb, s->sframe_lsp_bitsize);
1675 for (n = 0; n < MAX_FRAMES; n++) {
1676 int aw_idx_is_ext = 0;
1678 if (!s->has_residual_lsps) { // independent LSPs (per-frame)
1679 if (get_bits_left(gb) < s->frame_lsp_bitsize) return 1;
1680 skip_bits_long(gb, s->frame_lsp_bitsize);
1682 bd_idx = s->vbm_tree[get_vlc2(gb, frame_type_vlc.table, 6, 3)];
1684 return -1; // invalid frame type VLC code
1685 frame_desc = &frame_descs[bd_idx];
1686 if (frame_desc->acb_type == ACB_TYPE_ASYMMETRIC) {
1687 if (get_bits_left(gb) < s->pitch_nbits)
1689 skip_bits_long(gb, s->pitch_nbits);
1691 if (frame_desc->fcb_type == FCB_TYPE_SILENCE) {
1693 } else if (frame_desc->fcb_type == FCB_TYPE_AW_PULSES) {
1694 int tmp = get_bits(gb, 6);
1702 if (frame_desc->acb_type == ACB_TYPE_HAMMING) {
1703 need_bits = s->block_pitch_nbits +
1704 (frame_desc->n_blocks - 1) * s->block_delta_pitch_nbits;
1705 } else if (frame_desc->fcb_type == FCB_TYPE_AW_PULSES) {
1706 need_bits = 2 * !aw_idx_is_ext;
1709 need_bits += frame_desc->frame_size;
1710 if (get_bits_left(gb) < need_bits)
1712 skip_bits_long(gb, need_bits);
1719 * Synthesize output samples for a single superframe. If we have any data
1720 * cached in s->sframe_cache, that will be used instead of whatever is loaded
1723 * WMA Voice superframes contain 3 frames, each containing 160 audio samples,
1724 * to give a total of 480 samples per frame. See #synth_frame() for frame
1725 * parsing. In addition to 3 frames, superframes can also contain the LSPs
1726 * (if these are globally specified for all frames (residually); they can
1727 * also be specified individually per-frame. See the s->has_residual_lsps
1728 * option), and can specify the number of samples encoded in this superframe
1729 * (if less than 480), usually used to prevent blanks at track boundaries.
1731 * @param ctx WMA Voice decoder context
1732 * @return 0 on success, <0 on error or 1 if there was not enough data to
1733 * fully parse the superframe
1735 static int synth_superframe(AVCodecContext *ctx, int *got_frame_ptr)
1737 WMAVoiceContext *s = ctx->priv_data;
1738 GetBitContext *gb = &s->gb, s_gb;
1739 int n, res, n_samples = 480;
1740 double lsps[MAX_FRAMES][MAX_LSPS];
1741 const double *mean_lsf = s->lsps == 16 ?
1742 wmavoice_mean_lsf16[s->lsp_def_mode] : wmavoice_mean_lsf10[s->lsp_def_mode];
1743 float excitation[MAX_SIGNAL_HISTORY + MAX_SFRAMESIZE + 12];
1744 float synth[MAX_LSPS + MAX_SFRAMESIZE];
1747 memcpy(synth, s->synth_history,
1748 s->lsps * sizeof(*synth));
1749 memcpy(excitation, s->excitation_history,
1750 s->history_nsamples * sizeof(*excitation));
1752 if (s->sframe_cache_size > 0) {
1754 init_get_bits(gb, s->sframe_cache, s->sframe_cache_size);
1755 s->sframe_cache_size = 0;
1758 if ((res = check_bits_for_superframe(gb, s)) == 1) {
1763 /* First bit is speech/music bit, it differentiates between WMAVoice
1764 * speech samples (the actual codec) and WMAVoice music samples, which
1765 * are really WMAPro-in-WMAVoice-superframes. I've never seen those in
1767 if (!get_bits1(gb)) {
1768 av_log_missing_feature(ctx, "WMAPro-in-WMAVoice", 1);
1769 return AVERROR_PATCHWELCOME;
1772 /* (optional) nr. of samples in superframe; always <= 480 and >= 0 */
1773 if (get_bits1(gb)) {
1774 if ((n_samples = get_bits(gb, 12)) > 480) {
1775 av_log(ctx, AV_LOG_ERROR,
1776 "Superframe encodes >480 samples (%d), not allowed\n",
1781 /* Parse LSPs, if global for the superframe (can also be per-frame). */
1782 if (s->has_residual_lsps) {
1783 double prev_lsps[MAX_LSPS], a1[MAX_LSPS * 2], a2[MAX_LSPS * 2];
1785 for (n = 0; n < s->lsps; n++)
1786 prev_lsps[n] = s->prev_lsps[n] - mean_lsf[n];
1788 if (s->lsps == 10) {
1789 dequant_lsp10r(gb, lsps[2], prev_lsps, a1, a2, s->lsp_q_mode);
1790 } else /* s->lsps == 16 */
1791 dequant_lsp16r(gb, lsps[2], prev_lsps, a1, a2, s->lsp_q_mode);
1793 for (n = 0; n < s->lsps; n++) {
1794 lsps[0][n] = mean_lsf[n] + (a1[n] - a2[n * 2]);
1795 lsps[1][n] = mean_lsf[n] + (a1[s->lsps + n] - a2[n * 2 + 1]);
1796 lsps[2][n] += mean_lsf[n];
1798 for (n = 0; n < 3; n++)
1799 stabilize_lsps(lsps[n], s->lsps);
1802 /* get output buffer */
1803 s->frame.nb_samples = 480;
1804 if ((res = ctx->get_buffer(ctx, &s->frame)) < 0) {
1805 av_log(ctx, AV_LOG_ERROR, "get_buffer() failed\n");
1808 s->frame.nb_samples = n_samples;
1809 samples = (float *)s->frame.data[0];
1811 /* Parse frames, optionally preceded by per-frame (independent) LSPs. */
1812 for (n = 0; n < 3; n++) {
1813 if (!s->has_residual_lsps) {
1816 if (s->lsps == 10) {
1817 dequant_lsp10i(gb, lsps[n]);
1818 } else /* s->lsps == 16 */
1819 dequant_lsp16i(gb, lsps[n]);
1821 for (m = 0; m < s->lsps; m++)
1822 lsps[n][m] += mean_lsf[m];
1823 stabilize_lsps(lsps[n], s->lsps);
1826 if ((res = synth_frame(ctx, gb, n,
1827 &samples[n * MAX_FRAMESIZE],
1828 lsps[n], n == 0 ? s->prev_lsps : lsps[n - 1],
1829 &excitation[s->history_nsamples + n * MAX_FRAMESIZE],
1830 &synth[s->lsps + n * MAX_FRAMESIZE]))) {
1836 /* Statistics? FIXME - we don't check for length, a slight overrun
1837 * will be caught by internal buffer padding, and anything else
1838 * will be skipped, not read. */
1839 if (get_bits1(gb)) {
1840 res = get_bits(gb, 4);
1841 skip_bits(gb, 10 * (res + 1));
1846 /* Update history */
1847 memcpy(s->prev_lsps, lsps[2],
1848 s->lsps * sizeof(*s->prev_lsps));
1849 memcpy(s->synth_history, &synth[MAX_SFRAMESIZE],
1850 s->lsps * sizeof(*synth));
1851 memcpy(s->excitation_history, &excitation[MAX_SFRAMESIZE],
1852 s->history_nsamples * sizeof(*excitation));
1854 memmove(s->zero_exc_pf, &s->zero_exc_pf[MAX_SFRAMESIZE],
1855 s->history_nsamples * sizeof(*s->zero_exc_pf));
1861 * Parse the packet header at the start of each packet (input data to this
1864 * @param s WMA Voice decoding context private data
1865 * @return 1 if not enough bits were available, or 0 on success.
1867 static int parse_packet_header(WMAVoiceContext *s)
1869 GetBitContext *gb = &s->gb;
1872 if (get_bits_left(gb) < 11)
1874 skip_bits(gb, 4); // packet sequence number
1875 s->has_residual_lsps = get_bits1(gb);
1877 res = get_bits(gb, 6); // number of superframes per packet
1878 // (minus first one if there is spillover)
1879 if (get_bits_left(gb) < 6 * (res == 0x3F) + s->spillover_bitsize)
1881 } while (res == 0x3F);
1882 s->spillover_nbits = get_bits(gb, s->spillover_bitsize);
1888 * Copy (unaligned) bits from gb/data/size to pb.
1890 * @param pb target buffer to copy bits into
1891 * @param data source buffer to copy bits from
1892 * @param size size of the source data, in bytes
1893 * @param gb bit I/O context specifying the current position in the source.
1894 * data. This function might use this to align the bit position to
1895 * a whole-byte boundary before calling #avpriv_copy_bits() on aligned
1897 * @param nbits the amount of bits to copy from source to target
1899 * @note after calling this function, the current position in the input bit
1900 * I/O context is undefined.
1902 static void copy_bits(PutBitContext *pb,
1903 const uint8_t *data, int size,
1904 GetBitContext *gb, int nbits)
1906 int rmn_bytes, rmn_bits;
1908 rmn_bits = rmn_bytes = get_bits_left(gb);
1909 if (rmn_bits < nbits)
1911 if (nbits > pb->size_in_bits - put_bits_count(pb))
1913 rmn_bits &= 7; rmn_bytes >>= 3;
1914 if ((rmn_bits = FFMIN(rmn_bits, nbits)) > 0)
1915 put_bits(pb, rmn_bits, get_bits(gb, rmn_bits));
1916 avpriv_copy_bits(pb, data + size - rmn_bytes,
1917 FFMIN(nbits - rmn_bits, rmn_bytes << 3));
1921 * Packet decoding: a packet is anything that the (ASF) demuxer contains,
1922 * and we expect that the demuxer / application provides it to us as such
1923 * (else you'll probably get garbage as output). Every packet has a size of
1924 * ctx->block_align bytes, starts with a packet header (see
1925 * #parse_packet_header()), and then a series of superframes. Superframe
1926 * boundaries may exceed packets, i.e. superframes can split data over
1927 * multiple (two) packets.
1929 * For more information about frames, see #synth_superframe().
1931 static int wmavoice_decode_packet(AVCodecContext *ctx, void *data,
1932 int *got_frame_ptr, AVPacket *avpkt)
1934 WMAVoiceContext *s = ctx->priv_data;
1935 GetBitContext *gb = &s->gb;
1938 /* Packets are sometimes a multiple of ctx->block_align, with a packet
1939 * header at each ctx->block_align bytes. However, FFmpeg's ASF demuxer
1940 * feeds us ASF packets, which may concatenate multiple "codec" packets
1941 * in a single "muxer" packet, so we artificially emulate that by
1942 * capping the packet size at ctx->block_align. */
1943 for (size = avpkt->size; size > ctx->block_align; size -= ctx->block_align);
1948 init_get_bits(&s->gb, avpkt->data, size << 3);
1950 /* size == ctx->block_align is used to indicate whether we are dealing with
1951 * a new packet or a packet of which we already read the packet header
1953 if (size == ctx->block_align) { // new packet header
1954 if ((res = parse_packet_header(s)) < 0)
1957 /* If the packet header specifies a s->spillover_nbits, then we want
1958 * to push out all data of the previous packet (+ spillover) before
1959 * continuing to parse new superframes in the current packet. */
1960 if (s->spillover_nbits > 0) {
1961 if (s->sframe_cache_size > 0) {
1962 int cnt = get_bits_count(gb);
1963 copy_bits(&s->pb, avpkt->data, size, gb, s->spillover_nbits);
1964 flush_put_bits(&s->pb);
1965 s->sframe_cache_size += s->spillover_nbits;
1966 if ((res = synth_superframe(ctx, got_frame_ptr)) == 0 &&
1968 cnt += s->spillover_nbits;
1969 s->skip_bits_next = cnt & 7;
1970 *(AVFrame *)data = s->frame;
1973 skip_bits_long (gb, s->spillover_nbits - cnt +
1974 get_bits_count(gb)); // resync
1976 skip_bits_long(gb, s->spillover_nbits); // resync
1978 } else if (s->skip_bits_next)
1979 skip_bits(gb, s->skip_bits_next);
1981 /* Try parsing superframes in current packet */
1982 s->sframe_cache_size = 0;
1983 s->skip_bits_next = 0;
1984 pos = get_bits_left(gb);
1985 if ((res = synth_superframe(ctx, got_frame_ptr)) < 0) {
1987 } else if (*got_frame_ptr) {
1988 int cnt = get_bits_count(gb);
1989 s->skip_bits_next = cnt & 7;
1990 *(AVFrame *)data = s->frame;
1992 } else if ((s->sframe_cache_size = pos) > 0) {
1993 /* rewind bit reader to start of last (incomplete) superframe... */
1994 init_get_bits(gb, avpkt->data, size << 3);
1995 skip_bits_long(gb, (size << 3) - pos);
1996 av_assert1(get_bits_left(gb) == pos);
1998 /* ...and cache it for spillover in next packet */
1999 init_put_bits(&s->pb, s->sframe_cache, SFRAME_CACHE_MAXSIZE);
2000 copy_bits(&s->pb, avpkt->data, size, gb, s->sframe_cache_size);
2001 // FIXME bad - just copy bytes as whole and add use the
2002 // skip_bits_next field
2008 static av_cold int wmavoice_decode_end(AVCodecContext *ctx)
2010 WMAVoiceContext *s = ctx->priv_data;
2013 ff_rdft_end(&s->rdft);
2014 ff_rdft_end(&s->irdft);
2015 ff_dct_end(&s->dct);
2016 ff_dct_end(&s->dst);
2022 static av_cold void wmavoice_flush(AVCodecContext *ctx)
2024 WMAVoiceContext *s = ctx->priv_data;
2027 s->postfilter_agc = 0;
2028 s->sframe_cache_size = 0;
2029 s->skip_bits_next = 0;
2030 for (n = 0; n < s->lsps; n++)
2031 s->prev_lsps[n] = M_PI * (n + 1.0) / (s->lsps + 1.0);
2032 memset(s->excitation_history, 0,
2033 sizeof(*s->excitation_history) * MAX_SIGNAL_HISTORY);
2034 memset(s->synth_history, 0,
2035 sizeof(*s->synth_history) * MAX_LSPS);
2036 memset(s->gain_pred_err, 0,
2037 sizeof(s->gain_pred_err));
2040 memset(&s->synth_filter_out_buf[MAX_LSPS_ALIGN16 - s->lsps], 0,
2041 sizeof(*s->synth_filter_out_buf) * s->lsps);
2042 memset(s->dcf_mem, 0,
2043 sizeof(*s->dcf_mem) * 2);
2044 memset(s->zero_exc_pf, 0,
2045 sizeof(*s->zero_exc_pf) * s->history_nsamples);
2046 memset(s->denoise_filter_cache, 0, sizeof(s->denoise_filter_cache));
2050 AVCodec ff_wmavoice_decoder = {
2052 .type = AVMEDIA_TYPE_AUDIO,
2053 .id = AV_CODEC_ID_WMAVOICE,
2054 .priv_data_size = sizeof(WMAVoiceContext),
2055 .init = wmavoice_decode_init,
2056 .close = wmavoice_decode_end,
2057 .decode = wmavoice_decode_packet,
2058 .capabilities = CODEC_CAP_SUBFRAMES | CODEC_CAP_DR1,
2059 .flush = wmavoice_flush,
2060 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio Voice"),