yading@10
|
1 /*
|
yading@10
|
2 * AAC definitions and structures
|
yading@10
|
3 * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
|
yading@10
|
4 * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
|
yading@10
|
5 *
|
yading@10
|
6 * This file is part of FFmpeg.
|
yading@10
|
7 *
|
yading@10
|
8 * FFmpeg is free software; you can redistribute it and/or
|
yading@10
|
9 * modify it under the terms of the GNU Lesser General Public
|
yading@10
|
10 * License as published by the Free Software Foundation; either
|
yading@10
|
11 * version 2.1 of the License, or (at your option) any later version.
|
yading@10
|
12 *
|
yading@10
|
13 * FFmpeg is distributed in the hope that it will be useful,
|
yading@10
|
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
yading@10
|
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
yading@10
|
16 * Lesser General Public License for more details.
|
yading@10
|
17 *
|
yading@10
|
18 * You should have received a copy of the GNU Lesser General Public
|
yading@10
|
19 * License along with FFmpeg; if not, write to the Free Software
|
yading@10
|
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
yading@10
|
21 */
|
yading@10
|
22
|
yading@10
|
23 /**
|
yading@10
|
24 * @file
|
yading@10
|
25 * AAC definitions and structures
|
yading@10
|
26 * @author Oded Shimon ( ods15 ods15 dyndns org )
|
yading@10
|
27 * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
|
yading@10
|
28 */
|
yading@10
|
29
|
yading@10
|
30 #ifndef AVCODEC_AAC_H
|
yading@10
|
31 #define AVCODEC_AAC_H
|
yading@10
|
32
|
yading@10
|
33 #include "libavutil/float_dsp.h"
|
yading@10
|
34 #include "avcodec.h"
|
yading@10
|
35 #include "fft.h"
|
yading@10
|
36 #include "mpeg4audio.h"
|
yading@10
|
37 #include "sbr.h"
|
yading@10
|
38 #include "fmtconvert.h"
|
yading@10
|
39
|
yading@10
|
40 #include <stdint.h>
|
yading@10
|
41
|
yading@10
|
42 #define MAX_CHANNELS 64
|
yading@10
|
43 #define MAX_ELEM_ID 16
|
yading@10
|
44
|
yading@10
|
45 #define TNS_MAX_ORDER 20
|
yading@10
|
46 #define MAX_LTP_LONG_SFB 40
|
yading@10
|
47
|
yading@10
|
48 enum RawDataBlockType {
|
yading@10
|
49 TYPE_SCE,
|
yading@10
|
50 TYPE_CPE,
|
yading@10
|
51 TYPE_CCE,
|
yading@10
|
52 TYPE_LFE,
|
yading@10
|
53 TYPE_DSE,
|
yading@10
|
54 TYPE_PCE,
|
yading@10
|
55 TYPE_FIL,
|
yading@10
|
56 TYPE_END,
|
yading@10
|
57 };
|
yading@10
|
58
|
yading@10
|
59 enum ExtensionPayloadID {
|
yading@10
|
60 EXT_FILL,
|
yading@10
|
61 EXT_FILL_DATA,
|
yading@10
|
62 EXT_DATA_ELEMENT,
|
yading@10
|
63 EXT_DYNAMIC_RANGE = 0xb,
|
yading@10
|
64 EXT_SBR_DATA = 0xd,
|
yading@10
|
65 EXT_SBR_DATA_CRC = 0xe,
|
yading@10
|
66 };
|
yading@10
|
67
|
yading@10
|
68 enum WindowSequence {
|
yading@10
|
69 ONLY_LONG_SEQUENCE,
|
yading@10
|
70 LONG_START_SEQUENCE,
|
yading@10
|
71 EIGHT_SHORT_SEQUENCE,
|
yading@10
|
72 LONG_STOP_SEQUENCE,
|
yading@10
|
73 };
|
yading@10
|
74
|
yading@10
|
75 enum BandType {
|
yading@10
|
76 ZERO_BT = 0, ///< Scalefactors and spectral data are all zero.
|
yading@10
|
77 FIRST_PAIR_BT = 5, ///< This and later band types encode two values (rather than four) with one code word.
|
yading@10
|
78 ESC_BT = 11, ///< Spectral data are coded with an escape sequence.
|
yading@10
|
79 NOISE_BT = 13, ///< Spectral data are scaled white noise not coded in the bitstream.
|
yading@10
|
80 INTENSITY_BT2 = 14, ///< Scalefactor data are intensity stereo positions.
|
yading@10
|
81 INTENSITY_BT = 15, ///< Scalefactor data are intensity stereo positions.
|
yading@10
|
82 };
|
yading@10
|
83
|
yading@10
|
84 #define IS_CODEBOOK_UNSIGNED(x) ((x - 1) & 10)
|
yading@10
|
85
|
yading@10
|
86 enum ChannelPosition {
|
yading@10
|
87 AAC_CHANNEL_OFF = 0,
|
yading@10
|
88 AAC_CHANNEL_FRONT = 1,
|
yading@10
|
89 AAC_CHANNEL_SIDE = 2,
|
yading@10
|
90 AAC_CHANNEL_BACK = 3,
|
yading@10
|
91 AAC_CHANNEL_LFE = 4,
|
yading@10
|
92 AAC_CHANNEL_CC = 5,
|
yading@10
|
93 };
|
yading@10
|
94
|
yading@10
|
95 /**
|
yading@10
|
96 * The point during decoding at which channel coupling is applied.
|
yading@10
|
97 */
|
yading@10
|
98 enum CouplingPoint {
|
yading@10
|
99 BEFORE_TNS,
|
yading@10
|
100 BETWEEN_TNS_AND_IMDCT,
|
yading@10
|
101 AFTER_IMDCT = 3,
|
yading@10
|
102 };
|
yading@10
|
103
|
yading@10
|
104 /**
|
yading@10
|
105 * Output configuration status
|
yading@10
|
106 */
|
yading@10
|
107 enum OCStatus {
|
yading@10
|
108 OC_NONE, ///< Output unconfigured
|
yading@10
|
109 OC_TRIAL_PCE, ///< Output configuration under trial specified by an inband PCE
|
yading@10
|
110 OC_TRIAL_FRAME, ///< Output configuration under trial specified by a frame header
|
yading@10
|
111 OC_GLOBAL_HDR, ///< Output configuration set in a global header but not yet locked
|
yading@10
|
112 OC_LOCKED, ///< Output configuration locked in place
|
yading@10
|
113 };
|
yading@10
|
114
|
yading@10
|
115 typedef struct OutputConfiguration {
|
yading@10
|
116 MPEG4AudioConfig m4ac;
|
yading@10
|
117 uint8_t layout_map[MAX_ELEM_ID*4][3];
|
yading@10
|
118 int layout_map_tags;
|
yading@10
|
119 int channels;
|
yading@10
|
120 uint64_t channel_layout;
|
yading@10
|
121 enum OCStatus status;
|
yading@10
|
122 } OutputConfiguration;
|
yading@10
|
123
|
yading@10
|
124 /**
|
yading@10
|
125 * Predictor State
|
yading@10
|
126 */
|
yading@10
|
127 typedef struct PredictorState {
|
yading@10
|
128 float cor0;
|
yading@10
|
129 float cor1;
|
yading@10
|
130 float var0;
|
yading@10
|
131 float var1;
|
yading@10
|
132 float r0;
|
yading@10
|
133 float r1;
|
yading@10
|
134 } PredictorState;
|
yading@10
|
135
|
yading@10
|
136 #define MAX_PREDICTORS 672
|
yading@10
|
137
|
yading@10
|
138 #define SCALE_DIV_512 36 ///< scalefactor difference that corresponds to scale difference in 512 times
|
yading@10
|
139 #define SCALE_ONE_POS 140 ///< scalefactor index that corresponds to scale=1.0
|
yading@10
|
140 #define SCALE_MAX_POS 255 ///< scalefactor index maximum value
|
yading@10
|
141 #define SCALE_MAX_DIFF 60 ///< maximum scalefactor difference allowed by standard
|
yading@10
|
142 #define SCALE_DIFF_ZERO 60 ///< codebook index corresponding to zero scalefactor indices difference
|
yading@10
|
143
|
yading@10
|
144 /**
|
yading@10
|
145 * Long Term Prediction
|
yading@10
|
146 */
|
yading@10
|
147 typedef struct LongTermPrediction {
|
yading@10
|
148 int8_t present;
|
yading@10
|
149 int16_t lag;
|
yading@10
|
150 float coef;
|
yading@10
|
151 int8_t used[MAX_LTP_LONG_SFB];
|
yading@10
|
152 } LongTermPrediction;
|
yading@10
|
153
|
yading@10
|
154 /**
|
yading@10
|
155 * Individual Channel Stream
|
yading@10
|
156 */
|
yading@10
|
157 typedef struct IndividualChannelStream {
|
yading@10
|
158 uint8_t max_sfb; ///< number of scalefactor bands per group
|
yading@10
|
159 enum WindowSequence window_sequence[2];
|
yading@10
|
160 uint8_t use_kb_window[2]; ///< If set, use Kaiser-Bessel window, otherwise use a sinus window.
|
yading@10
|
161 int num_window_groups;
|
yading@10
|
162 uint8_t group_len[8];
|
yading@10
|
163 LongTermPrediction ltp;
|
yading@10
|
164 const uint16_t *swb_offset; ///< table of offsets to the lowest spectral coefficient of a scalefactor band, sfb, for a particular window
|
yading@10
|
165 const uint8_t *swb_sizes; ///< table of scalefactor band sizes for a particular window
|
yading@10
|
166 int num_swb; ///< number of scalefactor window bands
|
yading@10
|
167 int num_windows;
|
yading@10
|
168 int tns_max_bands;
|
yading@10
|
169 int predictor_present;
|
yading@10
|
170 int predictor_initialized;
|
yading@10
|
171 int predictor_reset_group;
|
yading@10
|
172 uint8_t prediction_used[41];
|
yading@10
|
173 } IndividualChannelStream;
|
yading@10
|
174
|
yading@10
|
175 /**
|
yading@10
|
176 * Temporal Noise Shaping
|
yading@10
|
177 */
|
yading@10
|
178 typedef struct TemporalNoiseShaping {
|
yading@10
|
179 int present;
|
yading@10
|
180 int n_filt[8];
|
yading@10
|
181 int length[8][4];
|
yading@10
|
182 int direction[8][4];
|
yading@10
|
183 int order[8][4];
|
yading@10
|
184 float coef[8][4][TNS_MAX_ORDER];
|
yading@10
|
185 } TemporalNoiseShaping;
|
yading@10
|
186
|
yading@10
|
187 /**
|
yading@10
|
188 * Dynamic Range Control - decoded from the bitstream but not processed further.
|
yading@10
|
189 */
|
yading@10
|
190 typedef struct DynamicRangeControl {
|
yading@10
|
191 int pce_instance_tag; ///< Indicates with which program the DRC info is associated.
|
yading@10
|
192 int dyn_rng_sgn[17]; ///< DRC sign information; 0 - positive, 1 - negative
|
yading@10
|
193 int dyn_rng_ctl[17]; ///< DRC magnitude information
|
yading@10
|
194 int exclude_mask[MAX_CHANNELS]; ///< Channels to be excluded from DRC processing.
|
yading@10
|
195 int band_incr; ///< Number of DRC bands greater than 1 having DRC info.
|
yading@10
|
196 int interpolation_scheme; ///< Indicates the interpolation scheme used in the SBR QMF domain.
|
yading@10
|
197 int band_top[17]; ///< Indicates the top of the i-th DRC band in units of 4 spectral lines.
|
yading@10
|
198 int prog_ref_level; /**< A reference level for the long-term program audio level for all
|
yading@10
|
199 * channels combined.
|
yading@10
|
200 */
|
yading@10
|
201 } DynamicRangeControl;
|
yading@10
|
202
|
yading@10
|
203 typedef struct Pulse {
|
yading@10
|
204 int num_pulse;
|
yading@10
|
205 int start;
|
yading@10
|
206 int pos[4];
|
yading@10
|
207 int amp[4];
|
yading@10
|
208 } Pulse;
|
yading@10
|
209
|
yading@10
|
210 /**
|
yading@10
|
211 * coupling parameters
|
yading@10
|
212 */
|
yading@10
|
213 typedef struct ChannelCoupling {
|
yading@10
|
214 enum CouplingPoint coupling_point; ///< The point during decoding at which coupling is applied.
|
yading@10
|
215 int num_coupled; ///< number of target elements
|
yading@10
|
216 enum RawDataBlockType type[8]; ///< Type of channel element to be coupled - SCE or CPE.
|
yading@10
|
217 int id_select[8]; ///< element id
|
yading@10
|
218 int ch_select[8]; /**< [0] shared list of gains; [1] list of gains for right channel;
|
yading@10
|
219 * [2] list of gains for left channel; [3] lists of gains for both channels
|
yading@10
|
220 */
|
yading@10
|
221 float gain[16][120];
|
yading@10
|
222 } ChannelCoupling;
|
yading@10
|
223
|
yading@10
|
224 /**
|
yading@10
|
225 * Single Channel Element - used for both SCE and LFE elements.
|
yading@10
|
226 */
|
yading@10
|
227 typedef struct SingleChannelElement {
|
yading@10
|
228 IndividualChannelStream ics;
|
yading@10
|
229 TemporalNoiseShaping tns;
|
yading@10
|
230 Pulse pulse;
|
yading@10
|
231 enum BandType band_type[128]; ///< band types
|
yading@10
|
232 int band_type_run_end[120]; ///< band type run end points
|
yading@10
|
233 float sf[120]; ///< scalefactors
|
yading@10
|
234 int sf_idx[128]; ///< scalefactor indices (used by encoder)
|
yading@10
|
235 uint8_t zeroes[128]; ///< band is not coded (used by encoder)
|
yading@10
|
236 DECLARE_ALIGNED(32, float, coeffs)[1024]; ///< coefficients for IMDCT
|
yading@10
|
237 DECLARE_ALIGNED(32, float, saved)[1024]; ///< overlap
|
yading@10
|
238 DECLARE_ALIGNED(32, float, ret_buf)[2048]; ///< PCM output buffer
|
yading@10
|
239 DECLARE_ALIGNED(16, float, ltp_state)[3072]; ///< time signal for LTP
|
yading@10
|
240 PredictorState predictor_state[MAX_PREDICTORS];
|
yading@10
|
241 float *ret; ///< PCM output
|
yading@10
|
242 } SingleChannelElement;
|
yading@10
|
243
|
yading@10
|
244 /**
|
yading@10
|
245 * channel element - generic struct for SCE/CPE/CCE/LFE
|
yading@10
|
246 */
|
yading@10
|
247 typedef struct ChannelElement {
|
yading@10
|
248 // CPE specific
|
yading@10
|
249 int common_window; ///< Set if channels share a common 'IndividualChannelStream' in bitstream.
|
yading@10
|
250 int ms_mode; ///< Signals mid/side stereo flags coding mode (used by encoder)
|
yading@10
|
251 uint8_t ms_mask[128]; ///< Set if mid/side stereo is used for each scalefactor window band
|
yading@10
|
252 // shared
|
yading@10
|
253 SingleChannelElement ch[2];
|
yading@10
|
254 // CCE specific
|
yading@10
|
255 ChannelCoupling coup;
|
yading@10
|
256 SpectralBandReplication sbr;
|
yading@10
|
257 } ChannelElement;
|
yading@10
|
258
|
yading@10
|
259 /**
|
yading@10
|
260 * main AAC context
|
yading@10
|
261 */
|
yading@10
|
262 struct AACContext {
|
yading@10
|
263 AVClass *class;
|
yading@10
|
264 AVCodecContext *avctx;
|
yading@10
|
265 AVFrame *frame;
|
yading@10
|
266
|
yading@10
|
267 int is_saved; ///< Set if elements have stored overlap from previous frame.
|
yading@10
|
268 DynamicRangeControl che_drc;
|
yading@10
|
269
|
yading@10
|
270 /**
|
yading@10
|
271 * @name Channel element related data
|
yading@10
|
272 * @{
|
yading@10
|
273 */
|
yading@10
|
274 ChannelElement *che[4][MAX_ELEM_ID];
|
yading@10
|
275 ChannelElement *tag_che_map[4][MAX_ELEM_ID];
|
yading@10
|
276 int tags_mapped;
|
yading@10
|
277 /** @} */
|
yading@10
|
278
|
yading@10
|
279 /**
|
yading@10
|
280 * @name temporary aligned temporary buffers
|
yading@10
|
281 * (We do not want to have these on the stack.)
|
yading@10
|
282 * @{
|
yading@10
|
283 */
|
yading@10
|
284 DECLARE_ALIGNED(32, float, buf_mdct)[1024];
|
yading@10
|
285 /** @} */
|
yading@10
|
286
|
yading@10
|
287 /**
|
yading@10
|
288 * @name Computed / set up during initialization
|
yading@10
|
289 * @{
|
yading@10
|
290 */
|
yading@10
|
291 FFTContext mdct;
|
yading@10
|
292 FFTContext mdct_small;
|
yading@10
|
293 FFTContext mdct_ltp;
|
yading@10
|
294 FmtConvertContext fmt_conv;
|
yading@10
|
295 AVFloatDSPContext fdsp;
|
yading@10
|
296 int random_state;
|
yading@10
|
297 /** @} */
|
yading@10
|
298
|
yading@10
|
299 /**
|
yading@10
|
300 * @name Members used for output
|
yading@10
|
301 * @{
|
yading@10
|
302 */
|
yading@10
|
303 SingleChannelElement *output_element[MAX_CHANNELS]; ///< Points to each SingleChannelElement
|
yading@10
|
304 /** @} */
|
yading@10
|
305
|
yading@10
|
306
|
yading@10
|
307 /**
|
yading@10
|
308 * @name Japanese DTV specific extension
|
yading@10
|
309 * @{
|
yading@10
|
310 */
|
yading@10
|
311 int force_dmono_mode;///< 0->not dmono, 1->use first channel, 2->use second channel
|
yading@10
|
312 int dmono_mode; ///< 0->not dmono, 1->use first channel, 2->use second channel
|
yading@10
|
313 /** @} */
|
yading@10
|
314
|
yading@10
|
315 DECLARE_ALIGNED(32, float, temp)[128];
|
yading@10
|
316
|
yading@10
|
317 OutputConfiguration oc[2];
|
yading@10
|
318 int warned_num_aac_frames;
|
yading@10
|
319
|
yading@10
|
320 /* aacdec functions pointers */
|
yading@10
|
321 void (*imdct_and_windowing)(AACContext *ac, SingleChannelElement *sce);
|
yading@10
|
322 void (*apply_ltp)(AACContext *ac, SingleChannelElement *sce);
|
yading@10
|
323 void (*apply_tns)(float coef[1024], TemporalNoiseShaping *tns,
|
yading@10
|
324 IndividualChannelStream *ics, int decode);
|
yading@10
|
325 void (*windowing_and_mdct_ltp)(AACContext *ac, float *out,
|
yading@10
|
326 float *in, IndividualChannelStream *ics);
|
yading@10
|
327 void (*update_ltp)(AACContext *ac, SingleChannelElement *sce);
|
yading@10
|
328
|
yading@10
|
329 };
|
yading@10
|
330
|
yading@10
|
331 void ff_aacdec_init_mips(AACContext *c);
|
yading@10
|
332
|
yading@10
|
333 #endif /* AVCODEC_AAC_H */
|