FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
ra288.c
Go to the documentation of this file.
1 /*
2  * RealAudio 2.0 (28.8K)
3  * Copyright (c) 2003 The FFmpeg Project
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 
23 #include "libavutil/float_dsp.h"
24 #include "libavutil/internal.h"
25 #include "avcodec.h"
26 #include "internal.h"
27 #define BITSTREAM_READER_LE
28 #include "get_bits.h"
29 #include "ra288.h"
30 #include "lpc.h"
31 #include "celp_filters.h"
32 
33 #define MAX_BACKWARD_FILTER_ORDER 36
34 #define MAX_BACKWARD_FILTER_LEN 40
35 #define MAX_BACKWARD_FILTER_NONREC 35
36 
37 #define RA288_BLOCK_SIZE 5
38 #define RA288_BLOCKS_PER_FRAME 32
39 
40 typedef struct RA288Context {
42  DECLARE_ALIGNED(32, float, sp_lpc)[FFALIGN(36, 16)]; ///< LPC coefficients for speech data (spec: A)
43  DECLARE_ALIGNED(32, float, gain_lpc)[FFALIGN(10, 16)]; ///< LPC coefficients for gain (spec: GB)
44 
45  /** speech data history (spec: SB).
46  * Its first 70 coefficients are updated only at backward filtering.
47  */
48  float sp_hist[111];
49 
50  /// speech part of the gain autocorrelation (spec: REXP)
51  float sp_rec[37];
52 
53  /** log-gain history (spec: SBLG).
54  * Its first 28 coefficients are updated only at backward filtering.
55  */
56  float gain_hist[38];
57 
58  /// recursive part of the gain autocorrelation (spec: REXPLG)
59  float gain_rec[11];
60 } RA288Context;
61 
63 {
64  RA288Context *ractx = avctx->priv_data;
65 
66  av_freep(&ractx->fdsp);
67 
68  return 0;
69 }
70 
72 {
73  RA288Context *ractx = avctx->priv_data;
74 
75  avctx->channels = 1;
78 
79  if (avctx->block_align <= 0) {
80  av_log(avctx, AV_LOG_ERROR, "unsupported block align\n");
81  return AVERROR_PATCHWELCOME;
82  }
83 
85  if (!ractx->fdsp)
86  return AVERROR(ENOMEM);
87 
88  return 0;
89 }
90 
91 static void convolve(float *tgt, const float *src, int len, int n)
92 {
93  for (; n >= 0; n--)
94  tgt[n] = avpriv_scalarproduct_float_c(src, src - n, len);
95 
96 }
97 
98 static void decode(RA288Context *ractx, float gain, int cb_coef)
99 {
100  int i;
101  double sumsum;
102  float sum, buffer[5];
103  float *block = ractx->sp_hist + 70 + 36; // current block
104  float *gain_block = ractx->gain_hist + 28;
105 
106  memmove(ractx->sp_hist + 70, ractx->sp_hist + 75, 36*sizeof(*block));
107 
108  /* block 46 of G.728 spec */
109  sum = 32.0;
110  for (i=0; i < 10; i++)
111  sum -= gain_block[9-i] * ractx->gain_lpc[i];
112 
113  /* block 47 of G.728 spec */
114  sum = av_clipf(sum, 0, 60);
115 
116  /* block 48 of G.728 spec */
117  /* exp(sum * 0.1151292546497) == pow(10.0,sum/20) */
118  sumsum = exp(sum * 0.1151292546497) * gain * (1.0/(1<<23));
119 
120  for (i=0; i < 5; i++)
121  buffer[i] = codetable[cb_coef][i] * sumsum;
122 
123  sum = avpriv_scalarproduct_float_c(buffer, buffer, 5);
124 
125  sum = FFMAX(sum, 5.0 / (1<<24));
126 
127  /* shift and store */
128  memmove(gain_block, gain_block + 1, 9 * sizeof(*gain_block));
129 
130  gain_block[9] = 10 * log10(sum) + (10*log10(((1<<24)/5.)) - 32);
131 
132  ff_celp_lp_synthesis_filterf(block, ractx->sp_lpc, buffer, 5, 36);
133 }
134 
135 /**
136  * Hybrid window filtering, see blocks 36 and 49 of the G.728 specification.
137  *
138  * @param order filter order
139  * @param n input length
140  * @param non_rec number of non-recursive samples
141  * @param out filter output
142  * @param hist pointer to the input history of the filter
143  * @param out pointer to the non-recursive part of the output
144  * @param out2 pointer to the recursive part of the output
145  * @param window pointer to the windowing function table
146  */
147 static void do_hybrid_window(RA288Context *ractx,
148  int order, int n, int non_rec, float *out,
149  float *hist, float *out2, const float *window)
150 {
151  int i;
152  float buffer1[MAX_BACKWARD_FILTER_ORDER + 1];
153  float buffer2[MAX_BACKWARD_FILTER_ORDER + 1];
157 
158  av_assert2(order>=0);
159 
160  ractx->fdsp->vector_fmul(work, window, hist, FFALIGN(order + n + non_rec, 16));
161 
162  convolve(buffer1, work + order , n , order);
163  convolve(buffer2, work + order + n, non_rec, order);
164 
165  for (i=0; i <= order; i++) {
166  out2[i] = out2[i] * 0.5625 + buffer1[i];
167  out [i] = out2[i] + buffer2[i];
168  }
169 
170  /* Multiply by the white noise correcting factor (WNCF). */
171  *out *= 257.0 / 256.0;
172 }
173 
174 /**
175  * Backward synthesis filter, find the LPC coefficients from past speech data.
176  */
177 static void backward_filter(RA288Context *ractx,
178  float *hist, float *rec, const float *window,
179  float *lpc, const float *tab,
180  int order, int n, int non_rec, int move_size)
181 {
183 
184  do_hybrid_window(ractx, order, n, non_rec, temp, hist, rec, window);
185 
186  if (!compute_lpc_coefs(temp, order, lpc, 0, 1, 1))
187  ractx->fdsp->vector_fmul(lpc, lpc, tab, FFALIGN(order, 16));
188 
189  memmove(hist, hist + n, move_size*sizeof(*hist));
190 }
191 
192 static int ra288_decode_frame(AVCodecContext * avctx, void *data,
193  int *got_frame_ptr, AVPacket *avpkt)
194 {
195  AVFrame *frame = data;
196  const uint8_t *buf = avpkt->data;
197  int buf_size = avpkt->size;
198  float *out;
199  int i, ret;
200  RA288Context *ractx = avctx->priv_data;
201  GetBitContext gb;
202 
203  if (buf_size < avctx->block_align) {
204  av_log(avctx, AV_LOG_ERROR,
205  "Error! Input buffer is too small [%d<%d]\n",
206  buf_size, avctx->block_align);
207  return AVERROR_INVALIDDATA;
208  }
209 
210  /* get output buffer */
212  if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
213  return ret;
214  out = (float *)frame->data[0];
215 
216  init_get_bits8(&gb, buf, avctx->block_align);
217 
218  for (i=0; i < RA288_BLOCKS_PER_FRAME; i++) {
219  float gain = amptable[get_bits(&gb, 3)];
220  int cb_coef = get_bits(&gb, 6 + (i&1));
221 
222  decode(ractx, gain, cb_coef);
223 
224  memcpy(out, &ractx->sp_hist[70 + 36], RA288_BLOCK_SIZE * sizeof(*out));
225  out += RA288_BLOCK_SIZE;
226 
227  if ((i & 7) == 3) {
228  backward_filter(ractx, ractx->sp_hist, ractx->sp_rec, syn_window,
229  ractx->sp_lpc, syn_bw_tab, 36, 40, 35, 70);
230 
231  backward_filter(ractx, ractx->gain_hist, ractx->gain_rec, gain_window,
232  ractx->gain_lpc, gain_bw_tab, 10, 8, 20, 28);
233  }
234  }
235 
236  *got_frame_ptr = 1;
237 
238  return avctx->block_align;
239 }
240 
242  .name = "real_288",
243  .long_name = NULL_IF_CONFIG_SMALL("RealAudio 2.0 (28.8K)"),
244  .type = AVMEDIA_TYPE_AUDIO,
245  .id = AV_CODEC_ID_RA_288,
246  .priv_data_size = sizeof(RA288Context),
249  .close = ra288_decode_close,
250  .capabilities = CODEC_CAP_DR1,
251 };
float sp_lpc[FFALIGN(36, 16)]
LPC coefficients for speech data (spec: A)
Definition: ra288.c:42
static av_cold int ra288_decode_close(AVCodecContext *avctx)
Definition: ra288.c:62
#define AVERROR_INVALIDDATA
Invalid data found when processing input.
Definition: error.h:59
void ff_celp_lp_synthesis_filterf(float *out, const float *filter_coeffs, const float *in, int buffer_length, int filter_length)
LP synthesis filter.
Definition: celp_filters.c:84
This structure describes decoded (raw) audio or video data.
Definition: frame.h:171
ptrdiff_t const GLvoid * data
Definition: opengl_enc.c:101
float gain_lpc[FFALIGN(10, 16)]
LPC coefficients for gain (spec: GB)
Definition: ra288.c:43
static void backward_filter(RA288Context *ractx, float *hist, float *rec, const float *window, float *lpc, const float *tab, int order, int n, int non_rec, int move_size)
Backward synthesis filter, find the LPC coefficients from past speech data.
Definition: ra288.c:177
static unsigned int get_bits(GetBitContext *s, int n)
Read 1-25 bits.
Definition: get_bits.h:260
else temp
Definition: vf_mcdeint.c:257
#define MAX_BACKWARD_FILTER_ORDER
Definition: ra288.c:33
static av_cold int init(AVCodecContext *avctx)
Definition: avrndec.c:35
int size
Definition: avcodec.h:1163
#define DECLARE_ALIGNED(n, t, v)
Definition: mem.h:53
static void decode(RA288Context *ractx, float gain, int cb_coef)
Definition: ra288.c:98
AVCodec.
Definition: avcodec.h:3181
int block_align
number of bytes per packet if constant and known or 0 Used by some WAV based audio codecs...
Definition: avcodec.h:2022
static const float amptable[8]
Definition: ra288.h:28
#define FFALIGN(x, a)
Definition: common.h:71
float avpriv_scalarproduct_float_c(const float *v1, const float *v2, int len)
Return the scalar product of two vectors.
Definition: float_dsp.c:108
float sp_hist[111]
speech data history (spec: SB).
Definition: ra288.c:48
enum AVSampleFormat sample_fmt
audio sample format
Definition: avcodec.h:1993
uint8_t
#define av_cold
Definition: attributes.h:74
#define MAX_BACKWARD_FILTER_NONREC
Definition: ra288.c:35
#define av_assert2(cond)
assert() equivalent, that does lie in speed critical code.
Definition: avassert.h:63
float gain_hist[38]
log-gain history (spec: SBLG).
Definition: ra288.c:56
#define CODEC_CAP_DR1
Codec uses get_buffer() for allocating buffers and supports custom allocators.
Definition: avcodec.h:789
static AVFrame * frame
uint8_t * data
Definition: avcodec.h:1162
bitstream reader API header.
#define CODEC_FLAG_BITEXACT
Use only bitexact stuff (except (I)DCT).
Definition: avcodec.h:759
#define av_log(a,...)
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:176
void(* vector_fmul)(float *dst, const float *src0, const float *src1, int len)
Calculate the entry wise product of two vectors of floats and store the result in a vector of floats...
Definition: float_dsp.h:38
#define AVERROR(e)
Definition: error.h:43
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:175
#define MAX_BACKWARD_FILTER_LEN
Definition: ra288.c:34
int flags
CODEC_FLAG_*.
Definition: avcodec.h:1335
const char * name
Name of the codec implementation.
Definition: avcodec.h:3188
static void convolve(float *tgt, const float *src, int len, int n)
Definition: ra288.c:91
#define FFMAX(a, b)
Definition: common.h:64
Libavcodec external API header.
static const float syn_window[FFALIGN(111, 16)]
Definition: ra288.h:100
uint64_t channel_layout
Audio channel layout.
Definition: avcodec.h:2046
AVCodec ff_ra_288_decoder
Definition: ra288.c:241
common internal API header
audio channel layout utility functions
#define RA288_BLOCKS_PER_FRAME
Definition: ra288.c:38
ret
Definition: avfilter.c:974
static const float gain_window[FFALIGN(38, 16)]
Definition: ra288.h:122
int n
Definition: avisynth_c.h:547
static const float syn_bw_tab[FFALIGN(36, 16)]
synthesis bandwidth broadening table
Definition: ra288.h:133
#define AVERROR_PATCHWELCOME
Not yet implemented in FFmpeg, patches welcome.
Definition: error.h:62
AVFloatDSPContext * fdsp
Definition: ra288.c:41
static const int16_t codetable[128][5]
Definition: ra288.h:33
AVS_Value src
Definition: avisynth_c.h:482
float sp_rec[37]
speech part of the gain autocorrelation (spec: REXP)
Definition: ra288.c:51
static int compute_lpc_coefs(const LPC_TYPE *autoc, int max_order, LPC_TYPE *lpc, int lpc_stride, int fail, int normalize)
Levinson-Durbin recursion.
Definition: lpc.h:155
static int init_get_bits8(GetBitContext *s, const uint8_t *buffer, int byte_size)
Initialize GetBitContext.
Definition: get_bits.h:441
main external API structure.
Definition: avcodec.h:1241
static void do_hybrid_window(RA288Context *ractx, int order, int n, int non_rec, float *out, float *hist, float *out2, const float *window)
Hybrid window filtering, see blocks 36 and 49 of the G.728 specification.
Definition: ra288.c:147
int ff_get_buffer(AVCodecContext *avctx, AVFrame *frame, int flags)
Get a buffer for a frame.
Definition: utils.c:1035
void * buf
Definition: avisynth_c.h:553
static const float gain_bw_tab[FFALIGN(10, 16)]
gain bandwidth broadening table
Definition: ra288.h:143
av_cold AVFloatDSPContext * avpriv_float_dsp_alloc(int bit_exact)
Allocate a float DSP context.
Definition: float_dsp.c:143
#define RA288_BLOCK_SIZE
Definition: ra288.c:37
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:182
static av_cold int ra288_decode_init(AVCodecContext *avctx)
Definition: ra288.c:71
static int ra288_decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, AVPacket *avpkt)
Definition: ra288.c:192
common internal api header.
#define LOCAL_ALIGNED(a, t, v,...)
Definition: internal.h:109
float gain_rec[11]
recursive part of the gain autocorrelation (spec: REXPLG)
Definition: ra288.c:59
void * priv_data
Definition: avcodec.h:1283
int len
int channels
number of audio channels
Definition: avcodec.h:1986
static const struct twinvq_data tab
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> out
#define av_freep(p)
#define AV_CH_LAYOUT_MONO
This structure stores compressed data.
Definition: avcodec.h:1139
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:225
GLuint buffer
Definition: opengl_enc.c:102
static int16_t block[64]
Definition: dct-test.c:110