libavcodec/ra288.c

   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
  22 #include "avcodec.h"
  23 #define ALT_BITSTREAM_READER_LE
  24 #include "bitstream.h"
  25 #include "ra288.h"
  26 #include "lpc.h"
  27 #include "celp_math.h"
  28
  29 typedef struct {
  30     float sp_lpc[36];      ///< LPC coefficients for speech data (spec: A)
  31     float gain_lpc[10];    ///< LPC coefficients for gain        (spec: GB)
  32
  33     /** speech data history                                      (spec: SB).
  34      *  Its first 70 coefficients are updated only at backward filtering.
  35      */
  36     float sp_hist[111];
  37
  38     /// speech part of the gain autocorrelation                  (spec: REXP)
  39     float sp_rec[37];
  40
  41     /** log-gain history                                         (spec: SBLG).
  42      *  Its first 28 coefficients are updated only at backward filtering.
  43      */
  44     float gain_hist[38];
  45
  46     /// recursive part of the gain autocorrelation               (spec: REXPLG)
  47     float gain_rec[11];
  48 } RA288Context;
  49
  50 static av_cold int ra288_decode_init(AVCodecContext *avctx)
  51 {
  52     avctx->sample_fmt = SAMPLE_FMT_FLT;
  53     return 0;
  54 }
  55
  56 static void apply_window(float *tgt, const float *m1, const float *m2, int n)
  57 {
  58     while (n--)
  59         *tgt++ = *m1++ * *m2++;
  60 }
  61
  62 static void convolve(float *tgt, const float *src, int len, int n)
  63 {
  64     for (; n >= 0; n--)
  65         tgt[n] = ff_dot_productf(src, src - n, len);
  66
  67 }
  68
  69 static void decode(RA288Context *ractx, float gain, int cb_coef)
  70 {
  71     int i, j;
  72     double sumsum;
  73     float sum, buffer[5];
  74     float *block = ractx->sp_hist + 70 + 36; // current block
  75     float *gain_block = ractx->gain_hist + 28;
  76
  77     memmove(ractx->sp_hist + 70, ractx->sp_hist + 75, 36*sizeof(*block));
  78
  79     /* block 46 of G.728 spec */
  80     sum = 32.;
  81     for (i=0; i < 10; i++)
  82         sum -= gain_block[9-i] * ractx->gain_lpc[i];
  83
  84     /* block 47 of G.728 spec */
  85     sum = av_clipf(sum, 0, 60);
  86
  87     /* block 48 of G.728 spec */
  88     /* exp(sum * 0.1151292546497) == pow(10.0,sum/20) */
  89     sumsum = exp(sum * 0.1151292546497) * gain * (1.0/(1<<23));
  90
  91     for (i=0; i < 5; i++)
  92         buffer[i] = codetable[cb_coef][i] * sumsum;
  93
  94     sum = ff_dot_productf(buffer, buffer, 5) * ((1<<24)/5.);
  95
  96     sum = FFMAX(sum, 1);
  97
  98     /* shift and store */
  99     memmove(gain_block, gain_block + 1, 9 * sizeof(*gain_block));
 100
 101     gain_block[9] = 10 * log10(sum) - 32;
 102
 103     for (i=0; i < 5; i++) {
 104         block[i] = buffer[i];
 105         for (j=0; j < 36; j++)
 106             block[i] -= block[i-1-j]*ractx->sp_lpc[j];
 107     }
 108
 109     /* output */
 110     for (i=0; i < 5; i++)
 111         block[i] = av_clipf(block[i], -4095./4096., 4095./4096.);
 112 }
 113
 114 /**
 115  * Hybrid window filtering, see blocks 36 and 49 of the G.728 specification.
 116  *
 117  * @param order   filter order
 118  * @param n       input length
 119  * @param non_rec number of non-recursive samples
 120  * @param out     filter output
 121  * @param hist    pointer to the input history of the filter
 122  * @param out     pointer to the non-recursive part of the output
 123  * @param out2    pointer to the recursive part of the output
 124  * @param window  pointer to the windowing function table
 125  */
 126 static void do_hybrid_window(int order, int n, int non_rec, float *out,
 127                              float *hist, float *out2, const float *window)
 128 {
 129     int i;
 130     float buffer1[order + 1];
 131     float buffer2[order + 1];
 132     float work[order + n + non_rec];
 133
 134     apply_window(work, window, hist, order + n + non_rec);
 135
 136     convolve(buffer1, work + order    , n      , order);
 137     convolve(buffer2, work + order + n, non_rec, order);
 138
 139     for (i=0; i <= order; i++) {
 140         out2[i] = out2[i] * 0.5625 + buffer1[i];
 141         out [i] = out2[i]          + buffer2[i];
 142     }
 143
 144     /* Multiply by the white noise correcting factor (WNCF). */
 145     *out *= 257./256.;
 146 }
 147
 148 /**
 149  * Backward synthesis filter, find the LPC coefficients from past speech data.
 150  */
 151 static void backward_filter(float *hist, float *rec, const float *window,
 152                             float *lpc, const float *tab,
 153                             int order, int n, int non_rec, int move_size)
 154 {
 155     float temp[order+1];
 156
 157     do_hybrid_window(order, n, non_rec, temp, hist, rec, window);
 158
 159     if (!compute_lpc_coefs(temp, order, lpc, 0, 1, 1))
 160         apply_window(lpc, lpc, tab, order);
 161
 162     memmove(hist, hist + n, move_size*sizeof(*hist));
 163 }
 164
 165 static int ra288_decode_frame(AVCodecContext * avctx, void *data,
 166                               int *data_size, const uint8_t * buf,
 167                               int buf_size)
 168 {
 169     float *out = data;
 170     int i, j;
 171     RA288Context *ractx = avctx->priv_data;
 172     GetBitContext gb;
 173
 174     if (buf_size < avctx->block_align) {
 175         av_log(avctx, AV_LOG_ERROR,
 176                "Error! Input buffer is too small [%d<%d]\n",
 177                buf_size, avctx->block_align);
 178         return 0;
 179     }
 180
 181     if (*data_size < 32*5*4)
 182         return -1;
 183
 184     init_get_bits(&gb, buf, avctx->block_align * 8);
 185
 186     for (i=0; i < 32; i++) {
 187         float gain = amptable[get_bits(&gb, 3)];
 188         int cb_coef = get_bits(&gb, 6 + (i&1));
 189
 190         decode(ractx, gain, cb_coef);
 191
 192         for (j=0; j < 5; j++)
 193             *(out++) = ractx->sp_hist[70 + 36 + j];
 194
 195         if ((i & 7) == 3) {
 196             backward_filter(ractx->sp_hist, ractx->sp_rec, syn_window,
 197                             ractx->sp_lpc, syn_bw_tab, 36, 40, 35, 70);
 198
 199             backward_filter(ractx->gain_hist, ractx->gain_rec, gain_window,
 200                             ractx->gain_lpc, gain_bw_tab, 10, 8, 20, 28);
 201         }
 202     }
 203
 204     *data_size = (char *)out - (char *)data;
 205     return avctx->block_align;
 206 }
 207
 208 AVCodec ra_288_decoder =
 209 {
 210     "real_288",
 211     CODEC_TYPE_AUDIO,
 212     CODEC_ID_RA_288,
 213     sizeof(RA288Context),
 214     ra288_decode_init,
 215     NULL,
 216     NULL,
 217     ra288_decode_frame,
 218     .long_name = NULL_IF_CONFIG_SMALL("RealAudio 2.0 (28.8K)"),
 219 };