FFmpeg
twinvq.h
Go to the documentation of this file.
1 /*
2  * TwinVQ decoder
3  * Copyright (c) 2009 Vitor Sessak
4  *
5  * This file is part of FFmpeg.
6  *
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.
11  *
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.
16  *
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
20  */
21 
22 #ifndef AVCODEC_TWINVQ_H
23 #define AVCODEC_TWINVQ_H
24 
25 #include <math.h>
26 #include <stdint.h>
27 
29 #include "libavutil/tx.h"
30 #include "libavutil/common.h"
31 #include "libavutil/float_dsp.h"
32 #include "avcodec.h"
33 
37 };
38 
40  TWINVQ_FT_SHORT = 0, ///< Short frame (divided in n sub-blocks)
41  TWINVQ_FT_MEDIUM, ///< Medium frame (divided in m<n sub-blocks)
42  TWINVQ_FT_LONG, ///< Long frame (single sub-block + PPC)
43  TWINVQ_FT_PPC, ///< Periodic Peak Component (part of the long frame)
44 };
45 
46 #define TWINVQ_PPC_SHAPE_CB_SIZE 64
47 #define TWINVQ_PPC_SHAPE_LEN_MAX 60
48 #define TWINVQ_SUB_AMP_MAX 4500.0
49 #define TWINVQ_MULAW_MU 100.0
50 #define TWINVQ_GAIN_BITS 8
51 #define TWINVQ_AMP_MAX 13000.0
52 #define TWINVQ_SUB_GAIN_BITS 5
53 #define TWINVQ_WINDOW_TYPE_BITS 4
54 #define TWINVQ_PGAIN_MU 200
55 #define TWINVQ_LSP_COEFS_MAX 20
56 #define TWINVQ_LSP_SPLIT_MAX 4
57 #define TWINVQ_CHANNELS_MAX 2
58 #define TWINVQ_SUBBLOCKS_MAX 16
59 #define TWINVQ_BARK_N_COEF_MAX 4
60 
61 #define TWINVQ_MAX_FRAMES_PER_PACKET 2
62 
63 /**
64  * Parameters and tables that are different for each frame type
65  */
67  uint8_t sub; ///< Number subblocks in each frame
68  const uint16_t *bark_tab;
69 
70  /** number of distinct bark scale envelope values */
71  uint8_t bark_env_size;
72 
73  const int16_t *bark_cb; ///< codebook for the bark scale envelope (BSE)
74  uint8_t bark_n_coef;///< number of BSE CB coefficients to read
75  uint8_t bark_n_bit; ///< number of bits of the BSE coefs
76 
77  //@{
78  /** main codebooks for spectrum data */
79  const int16_t *cb0;
80  const int16_t *cb1;
81  //@}
82 
83  uint8_t cb_len_read; ///< number of spectrum coefficients to read
84 };
85 
86 typedef struct TwinVQFrameData {
89 
90  uint8_t main_coeffs[1024];
92 
95 
98 
102 
106 
107 /**
108  * Parameters and tables that are different for every combination of
109  * bitrate/sample rate
110  */
111 typedef struct TwinVQModeTab {
112  struct TwinVQFrameMode fmode[3]; ///< frame type-dependent parameters
113 
114  uint16_t size; ///< frame size in samples
115  uint8_t n_lsp; ///< number of lsp coefficients
116  const float *lspcodebook;
117 
118  /* number of bits of the different LSP CB coefficients */
119  uint8_t lsp_bit0;
120  uint8_t lsp_bit1;
121  uint8_t lsp_bit2;
122 
123  uint8_t lsp_split; ///< number of CB entries for the LSP decoding
124  const int16_t *ppc_shape_cb; ///< PPC shape CB
125 
126  /** number of the bits for the PPC period value */
127  uint8_t ppc_period_bit;
128 
129  uint8_t ppc_shape_bit; ///< number of bits of the PPC shape CB coeffs
130  uint8_t ppc_shape_len; ///< size of PPC shape CB
131  uint8_t pgain_bit; ///< bits for PPC gain
132 
133  /** constant for peak period to peak width conversion */
134  uint16_t peak_per2wid;
135 } TwinVQModeTab;
136 
137 typedef struct TwinVQContext {
142 
144 
145  int is_6kbps;
146 
147  // history
148  float lsp_hist[2][20]; ///< LSP coefficients of the last frame
149  float bark_hist[3][2][40]; ///< BSE coefficients of last frame
150 
151  // bitstream parameters
152  int16_t permut[4][4096];
153  uint8_t length[4][2]; ///< main codebook stride
154  uint8_t length_change[4];
155  uint8_t bits_main_spec[2][4][2]; ///< bits for the main codebook
157  int n_div[4];
158 
159  float *spectrum;
160  float *curr_frame; ///< non-interleaved output
161  float *prev_frame; ///< non-interleaved previous frame
164 
165  float *cos_tabs[3];
166 
167  // scratch buffers
168  float *tmp_buf;
169 
172 
174 
176  const uint8_t *buf, int buf_size);
177  void (*dec_bark_env)(struct TwinVQContext *tctx, const uint8_t *in,
178  int use_hist, int ch, float *out, float gain,
179  enum TwinVQFrameType ftype);
180  void (*decode_ppc)(struct TwinVQContext *tctx, int period_coef, int g_coef,
181  const float *shape, float *speech);
182 } TwinVQContext;
183 
186 
187 extern const float ff_metasound_lsp8[];
188 extern const float ff_metasound_lsp11[];
189 extern const float ff_metasound_lsp16[];
190 extern const float ff_metasound_lsp22[];
191 extern const float ff_metasound_lsp44[];
192 
193 /** @note not speed critical, hence not optimized */
194 static inline void twinvq_memset_float(float *buf, float val, int size)
195 {
196  while (size--)
197  *buf++ = val;
198 }
199 
200 static inline float twinvq_mulawinv(float y, float clip, float mu)
201 {
202  y = av_clipf(y / clip, -1, 1);
203  return clip * FFSIGN(y) * (exp(log(1 + mu) * fabs(y)) - 1) / mu;
204 }
205 
207  int *got_frame_ptr, AVPacket *avpkt);
209 /** Requires the caller to call ff_twinvq_decode_close() upon failure. */
212 
213 #endif /* AVCODEC_TWINVQ_H */
TwinVQModeTab::lspcodebook
const float * lspcodebook
Definition: twinvq.h:116
FF_VISIBILITY_PUSH_HIDDEN
#define FF_VISIBILITY_PUSH_HIDDEN
Definition: attributes_internal.h:30
TwinVQContext::mtab
const TwinVQModeTab * mtab
Definition: twinvq.h:143
TwinVQFrameMode::sub
uint8_t sub
Number subblocks in each frame.
Definition: twinvq.h:67
out
FILE * out
Definition: movenc.c:54
TwinVQContext::bits_main_spec
uint8_t bits_main_spec[2][4][2]
bits for the main codebook
Definition: twinvq.h:155
TwinVQFrameData::bark1
uint8_t bark1[TWINVQ_CHANNELS_MAX][TWINVQ_SUBBLOCKS_MAX][TWINVQ_BARK_N_COEF_MAX]
Definition: twinvq.h:96
TwinVQFrameData::lpc_idx2
uint8_t lpc_idx2[TWINVQ_CHANNELS_MAX][TWINVQ_LSP_SPLIT_MAX]
Definition: twinvq.h:100
TwinVQContext::tmp_buf
float * tmp_buf
Definition: twinvq.h:168
TwinVQContext::curr_frame
float * curr_frame
non-interleaved output
Definition: twinvq.h:160
AVTXContext
Definition: tx_priv.h:235
TwinVQFrameData::bark_use_hist
uint8_t bark_use_hist[TWINVQ_CHANNELS_MAX][TWINVQ_SUBBLOCKS_MAX]
Definition: twinvq.h:97
normalize.log
log
Definition: normalize.py:21
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:340
TwinVQContext::bits
TwinVQFrameData bits[TWINVQ_MAX_FRAMES_PER_PACKET]
Definition: twinvq.h:171
TwinVQModeTab::ppc_shape_len
uint8_t ppc_shape_len
size of PPC shape CB
Definition: twinvq.h:130
TwinVQModeTab::lsp_bit1
uint8_t lsp_bit1
Definition: twinvq.h:120
TwinVQContext::permut
int16_t permut[4][4096]
Definition: twinvq.h:152
TwinVQModeTab::pgain_bit
uint8_t pgain_bit
bits for PPC gain
Definition: twinvq.h:131
TWINVQ_CODEC_VQF
@ TWINVQ_CODEC_VQF
Definition: twinvq.h:35
TwinVQFrameData::ppc_coeffs
uint8_t ppc_coeffs[TWINVQ_PPC_SHAPE_LEN_MAX]
Definition: twinvq.h:91
TwinVQFrameMode::cb0
const int16_t * cb0
main codebooks for spectrum data
Definition: twinvq.h:79
TwinVQContext::last_block_pos
int last_block_pos[2]
Definition: twinvq.h:162
TwinVQFrameMode::cb1
const int16_t * cb1
Definition: twinvq.h:80
TwinVQFrameMode::bark_n_coef
uint8_t bark_n_coef
number of BSE CB coefficients to read
Definition: twinvq.h:74
TwinVQFrameType
TwinVQFrameType
Definition: twinvq.h:39
FFSIGN
#define FFSIGN(a)
Definition: common.h:66
ff_metasound_lsp16
const float ff_metasound_lsp16[]
Definition: metasound_twinvq_data.h:260
ff_metasound_lsp22
const float ff_metasound_lsp22[]
Definition: metasound_twinvq_data.h:439
TwinVQFrameData::sub_gain_bits
uint8_t sub_gain_bits[TWINVQ_CHANNELS_MAX *TWINVQ_SUBBLOCKS_MAX]
Definition: twinvq.h:94
val
static double val(void *priv, double ch)
Definition: aeval.c:78
TwinVQModeTab::size
uint16_t size
frame size in samples
Definition: twinvq.h:114
twinvq_mulawinv
static float twinvq_mulawinv(float y, float clip, float mu)
Definition: twinvq.h:200
TwinVQContext::dec_bark_env
void(* dec_bark_env)(struct TwinVQContext *tctx, const uint8_t *in, int use_hist, int ch, float *out, float gain, enum TwinVQFrameType ftype)
Definition: twinvq.h:177
TwinVQContext::n_div
int n_div[4]
Definition: twinvq.h:157
TwinVQContext::decode_ppc
void(* decode_ppc)(struct TwinVQContext *tctx, int period_coef, int g_coef, const float *shape, float *speech)
Definition: twinvq.h:180
TwinVQContext::fdsp
AVFloatDSPContext * fdsp
Definition: twinvq.h:139
TwinVQFrameData::g_coef
int g_coef[TWINVQ_CHANNELS_MAX]
Definition: twinvq.h:104
ff_twinvq_decode_init
int ff_twinvq_decode_init(AVCodecContext *avctx)
Requires the caller to call ff_twinvq_decode_close() upon failure.
Definition: twinvq.c:763
av_tx_fn
void(* av_tx_fn)(AVTXContext *s, void *out, void *in, ptrdiff_t stride)
Function pointer to a function to perform the transform.
Definition: tx.h:151
TwinVQContext::lsp_hist
float lsp_hist[2][20]
LSP coefficients of the last frame.
Definition: twinvq.h:148
clip
clip
Definition: af_crystalizer.c:121
TwinVQModeTab::ppc_shape_bit
uint8_t ppc_shape_bit
number of bits of the PPC shape CB coeffs
Definition: twinvq.h:129
TwinVQFrameData::lpc_idx1
uint8_t lpc_idx1[TWINVQ_CHANNELS_MAX]
Definition: twinvq.h:99
attributes_internal.h
TWINVQ_BARK_N_COEF_MAX
#define TWINVQ_BARK_N_COEF_MAX
Definition: twinvq.h:59
TwinVQContext::codec
enum TwinVQCodec codec
Definition: twinvq.h:173
TwinVQContext::length_change
uint8_t length_change[4]
Definition: twinvq.h:154
TwinVQFrameData::lpc_hist_idx
uint8_t lpc_hist_idx[TWINVQ_CHANNELS_MAX]
Definition: twinvq.h:101
ff_twinvq_decode_frame
int ff_twinvq_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_frame_ptr, AVPacket *avpkt)
Definition: twinvq.c:479
TWINVQ_FT_MEDIUM
@ TWINVQ_FT_MEDIUM
Medium frame (divided in m<n sub-blocks)
Definition: twinvq.h:41
TwinVQFrameMode
Parameters and tables that are different for each frame type.
Definition: twinvq.h:66
TwinVQModeTab
Parameters and tables that are different for every combination of bitrate/sample rate.
Definition: twinvq.h:111
TwinVQModeTab::lsp_bit2
uint8_t lsp_bit2
Definition: twinvq.h:121
TwinVQFrameMode::bark_tab
const uint16_t * bark_tab
Definition: twinvq.h:68
twinvq_memset_float
static void twinvq_memset_float(float *buf, float val, int size)
Definition: twinvq.h:194
TwinVQContext::prev_frame
float * prev_frame
non-interleaved previous frame
Definition: twinvq.h:161
frame
static AVFrame * frame
Definition: demux_decode.c:54
TWINVQ_FT_PPC
@ TWINVQ_FT_PPC
Periodic Peak Component (part of the long frame)
Definition: twinvq.h:43
TwinVQFrameMode::bark_cb
const int16_t * bark_cb
codebook for the bark scale envelope (BSE)
Definition: twinvq.h:73
TwinVQContext::length
uint8_t length[4][2]
main codebook stride
Definition: twinvq.h:153
TwinVQFrameData::gain_bits
uint8_t gain_bits[TWINVQ_CHANNELS_MAX]
Definition: twinvq.h:93
fabs
static __device__ float fabs(float a)
Definition: cuda_runtime.h:182
av_clipf
av_clipf
Definition: af_crystalizer.c:121
TwinVQContext::is_6kbps
int is_6kbps
Definition: twinvq.h:145
exp
int8_t exp
Definition: eval.c:72
float_dsp.h
TwinVQFrameData::main_coeffs
uint8_t main_coeffs[1024]
Definition: twinvq.h:90
TwinVQContext::avctx
AVCodecContext * avctx
Definition: twinvq.h:138
TwinVQModeTab::lsp_bit0
uint8_t lsp_bit0
Definition: twinvq.h:119
TWINVQ_FT_LONG
@ TWINVQ_FT_LONG
Long frame (single sub-block + PPC)
Definition: twinvq.h:42
ff_metasound_lsp44
const float ff_metasound_lsp44[]
Definition: metasound_twinvq_data.h:606
TwinVQFrameMode::bark_n_bit
uint8_t bark_n_bit
number of bits of the BSE coefs
Definition: twinvq.h:75
size
int size
Definition: twinvq_data.h:10344
TwinVQModeTab::ppc_period_bit
uint8_t ppc_period_bit
number of the bits for the PPC period value
Definition: twinvq.h:127
TwinVQContext::bits_main_spec_change
int bits_main_spec_change[4]
Definition: twinvq.h:156
TWINVQ_PPC_SHAPE_LEN_MAX
#define TWINVQ_PPC_SHAPE_LEN_MAX
Definition: twinvq.h:47
TWINVQ_CHANNELS_MAX
#define TWINVQ_CHANNELS_MAX
Definition: twinvq.h:57
TwinVQModeTab::ppc_shape_cb
const int16_t * ppc_shape_cb
PPC shape CB.
Definition: twinvq.h:124
AVFloatDSPContext
Definition: float_dsp.h:24
TwinVQModeTab::peak_per2wid
uint16_t peak_per2wid
constant for peak period to peak width conversion
Definition: twinvq.h:134
FF_VISIBILITY_POP_HIDDEN
#define FF_VISIBILITY_POP_HIDDEN
Definition: attributes_internal.h:31
TwinVQContext::read_bitstream
int(* read_bitstream)(AVCodecContext *avctx, struct TwinVQContext *tctx, const uint8_t *buf, int buf_size)
Definition: twinvq.h:175
TwinVQContext::discarded_packets
int discarded_packets
Definition: twinvq.h:163
TWINVQ_CODEC_METASOUND
@ TWINVQ_CODEC_METASOUND
Definition: twinvq.h:36
TwinVQContext::frame_size
int frame_size
Definition: twinvq.h:170
TwinVQContext::bark_hist
float bark_hist[3][2][40]
BSE coefficients of last frame.
Definition: twinvq.h:149
TwinVQFrameData
Definition: twinvq.h:86
common.h
TwinVQContext::cos_tabs
float * cos_tabs[3]
Definition: twinvq.h:165
ff_twinvq_decode_close
int ff_twinvq_decode_close(AVCodecContext *avctx)
Definition: twinvq.c:744
ff_metasound_lsp8
const float ff_metasound_lsp8[]
Definition: metasound_twinvq_data.h:27
avcodec.h
TwinVQContext::cur_frame
int cur_frame
Definition: twinvq.h:170
TwinVQContext::tx_fn
av_tx_fn tx_fn[3]
Definition: twinvq.h:141
TWINVQ_FT_SHORT
@ TWINVQ_FT_SHORT
Short frame (divided in n sub-blocks)
Definition: twinvq.h:40
AVCodecContext
main external API structure.
Definition: avcodec.h:441
ff_metasound_lsp11
const float ff_metasound_lsp11[]
Definition: metasound_twinvq_data.h:93
TwinVQFrameMode::cb_len_read
uint8_t cb_len_read
number of spectrum coefficients to read
Definition: twinvq.h:83
TwinVQFrameData::ftype
enum TwinVQFrameType ftype
Definition: twinvq.h:88
ff_twinvq_wtype_to_ftype_table
FF_VISIBILITY_PUSH_HIDDEN enum TwinVQFrameType ff_twinvq_wtype_to_ftype_table[]
Definition: twinvq.c:473
TwinVQContext::tx
AVTXContext * tx[3]
Definition: twinvq.h:140
TwinVQContext::frames_per_packet
int frames_per_packet
Definition: twinvq.h:170
AVPacket
This structure stores compressed data.
Definition: packet.h:468
ftype
#define ftype
Definition: adynamicequalizer_template.c:55
TWINVQ_SUBBLOCKS_MAX
#define TWINVQ_SUBBLOCKS_MAX
Definition: twinvq.h:58
TwinVQContext::spectrum
float * spectrum
Definition: twinvq.h:159
TwinVQFrameData::window_type
int window_type
Definition: twinvq.h:87
TWINVQ_MAX_FRAMES_PER_PACKET
#define TWINVQ_MAX_FRAMES_PER_PACKET
Definition: twinvq.h:61
TwinVQFrameData::p_coef
int p_coef[TWINVQ_CHANNELS_MAX]
Definition: twinvq.h:103
TwinVQContext
Definition: twinvq.h:137
TWINVQ_LSP_SPLIT_MAX
#define TWINVQ_LSP_SPLIT_MAX
Definition: twinvq.h:56
int
int
Definition: ffmpeg_filter.c:368
TwinVQModeTab::lsp_split
uint8_t lsp_split
number of CB entries for the LSP decoding
Definition: twinvq.h:123
TwinVQModeTab::fmode
struct TwinVQFrameMode fmode[3]
frame type-dependent parameters
Definition: twinvq.h:112
TwinVQFrameMode::bark_env_size
uint8_t bark_env_size
number of distinct bark scale envelope values
Definition: twinvq.h:71
tx.h
TwinVQModeTab::n_lsp
uint8_t n_lsp
number of lsp coefficients
Definition: twinvq.h:115
TwinVQCodec
TwinVQCodec
Definition: twinvq.h:34