You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

255 lines
8.0KB

  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. #include "libavutil/channel_layout.h"
  22. #include "libavutil/float_dsp.h"
  23. #include "libavutil/internal.h"
  24. #define BITSTREAM_READER_LE
  25. #include "avcodec.h"
  26. #include "celp_filters.h"
  27. #include "get_bits.h"
  28. #include "internal.h"
  29. #include "lpc.h"
  30. #include "ra288.h"
  31. #define MAX_BACKWARD_FILTER_ORDER 36
  32. #define MAX_BACKWARD_FILTER_LEN 40
  33. #define MAX_BACKWARD_FILTER_NONREC 35
  34. #define RA288_BLOCK_SIZE 5
  35. #define RA288_BLOCKS_PER_FRAME 32
  36. typedef struct RA288Context {
  37. AVFloatDSPContext *fdsp;
  38. DECLARE_ALIGNED(32, float, sp_lpc)[FFALIGN(36, 16)]; ///< LPC coefficients for speech data (spec: A)
  39. DECLARE_ALIGNED(32, float, gain_lpc)[FFALIGN(10, 16)]; ///< LPC coefficients for gain (spec: GB)
  40. /** speech data history (spec: SB).
  41. * Its first 70 coefficients are updated only at backward filtering.
  42. */
  43. float sp_hist[111];
  44. /// speech part of the gain autocorrelation (spec: REXP)
  45. float sp_rec[37];
  46. /** log-gain history (spec: SBLG).
  47. * Its first 28 coefficients are updated only at backward filtering.
  48. */
  49. float gain_hist[38];
  50. /// recursive part of the gain autocorrelation (spec: REXPLG)
  51. float gain_rec[11];
  52. } RA288Context;
  53. static av_cold int ra288_decode_close(AVCodecContext *avctx)
  54. {
  55. RA288Context *ractx = avctx->priv_data;
  56. av_freep(&ractx->fdsp);
  57. return 0;
  58. }
  59. static av_cold int ra288_decode_init(AVCodecContext *avctx)
  60. {
  61. RA288Context *ractx = avctx->priv_data;
  62. avctx->channels = 1;
  63. avctx->channel_layout = AV_CH_LAYOUT_MONO;
  64. avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
  65. if (avctx->block_align <= 0) {
  66. av_log(avctx, AV_LOG_ERROR, "unsupported block align\n");
  67. return AVERROR_PATCHWELCOME;
  68. }
  69. ractx->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
  70. if (!ractx->fdsp)
  71. return AVERROR(ENOMEM);
  72. return 0;
  73. }
  74. static void convolve(float *tgt, const float *src, int len, int n)
  75. {
  76. for (; n >= 0; n--)
  77. tgt[n] = avpriv_scalarproduct_float_c(src, src - n, len);
  78. }
  79. static void decode(RA288Context *ractx, float gain, int cb_coef)
  80. {
  81. int i;
  82. double sumsum;
  83. float sum, buffer[5];
  84. float *block = ractx->sp_hist + 70 + 36; // current block
  85. float *gain_block = ractx->gain_hist + 28;
  86. memmove(ractx->sp_hist + 70, ractx->sp_hist + 75, 36*sizeof(*block));
  87. /* block 46 of G.728 spec */
  88. sum = 32.0;
  89. for (i=0; i < 10; i++)
  90. sum -= gain_block[9-i] * ractx->gain_lpc[i];
  91. /* block 47 of G.728 spec */
  92. sum = av_clipf(sum, 0, 60);
  93. /* block 48 of G.728 spec */
  94. /* exp(sum * 0.1151292546497) == pow(10.0,sum/20) */
  95. sumsum = exp(sum * 0.1151292546497) * gain * (1.0/(1<<23));
  96. for (i=0; i < 5; i++)
  97. buffer[i] = codetable[cb_coef][i] * sumsum;
  98. sum = avpriv_scalarproduct_float_c(buffer, buffer, 5);
  99. sum = FFMAX(sum, 5.0 / (1<<24));
  100. /* shift and store */
  101. memmove(gain_block, gain_block + 1, 9 * sizeof(*gain_block));
  102. gain_block[9] = 10 * log10(sum) + (10*log10(((1<<24)/5.)) - 32);
  103. ff_celp_lp_synthesis_filterf(block, ractx->sp_lpc, buffer, 5, 36);
  104. }
  105. /**
  106. * Hybrid window filtering, see blocks 36 and 49 of the G.728 specification.
  107. *
  108. * @param order filter order
  109. * @param n input length
  110. * @param non_rec number of non-recursive samples
  111. * @param out filter output
  112. * @param hist pointer to the input history of the filter
  113. * @param out pointer to the non-recursive part of the output
  114. * @param out2 pointer to the recursive part of the output
  115. * @param window pointer to the windowing function table
  116. */
  117. static void do_hybrid_window(RA288Context *ractx,
  118. int order, int n, int non_rec, float *out,
  119. float *hist, float *out2, const float *window)
  120. {
  121. int i;
  122. float buffer1[MAX_BACKWARD_FILTER_ORDER + 1];
  123. float buffer2[MAX_BACKWARD_FILTER_ORDER + 1];
  124. LOCAL_ALIGNED(32, float, work, [FFALIGN(MAX_BACKWARD_FILTER_ORDER +
  125. MAX_BACKWARD_FILTER_LEN +
  126. MAX_BACKWARD_FILTER_NONREC, 16)]);
  127. av_assert2(order>=0);
  128. ractx->fdsp->vector_fmul(work, window, hist, FFALIGN(order + n + non_rec, 16));
  129. convolve(buffer1, work + order , n , order);
  130. convolve(buffer2, work + order + n, non_rec, order);
  131. for (i=0; i <= order; i++) {
  132. out2[i] = out2[i] * 0.5625 + buffer1[i];
  133. out [i] = out2[i] + buffer2[i];
  134. }
  135. /* Multiply by the white noise correcting factor (WNCF). */
  136. *out *= 257.0 / 256.0;
  137. }
  138. /**
  139. * Backward synthesis filter, find the LPC coefficients from past speech data.
  140. */
  141. static void backward_filter(RA288Context *ractx,
  142. float *hist, float *rec, const float *window,
  143. float *lpc, const float *tab,
  144. int order, int n, int non_rec, int move_size)
  145. {
  146. float temp[MAX_BACKWARD_FILTER_ORDER+1];
  147. do_hybrid_window(ractx, order, n, non_rec, temp, hist, rec, window);
  148. if (!compute_lpc_coefs(temp, order, lpc, 0, 1, 1))
  149. ractx->fdsp->vector_fmul(lpc, lpc, tab, FFALIGN(order, 16));
  150. memmove(hist, hist + n, move_size*sizeof(*hist));
  151. }
  152. static int ra288_decode_frame(AVCodecContext * avctx, void *data,
  153. int *got_frame_ptr, AVPacket *avpkt)
  154. {
  155. AVFrame *frame = data;
  156. const uint8_t *buf = avpkt->data;
  157. int buf_size = avpkt->size;
  158. float *out;
  159. int i, ret;
  160. RA288Context *ractx = avctx->priv_data;
  161. GetBitContext gb;
  162. if (buf_size < avctx->block_align) {
  163. av_log(avctx, AV_LOG_ERROR,
  164. "Error! Input buffer is too small [%d<%d]\n",
  165. buf_size, avctx->block_align);
  166. return AVERROR_INVALIDDATA;
  167. }
  168. ret = init_get_bits8(&gb, buf, avctx->block_align);
  169. if (ret < 0)
  170. return ret;
  171. /* get output buffer */
  172. frame->nb_samples = RA288_BLOCK_SIZE * RA288_BLOCKS_PER_FRAME;
  173. if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
  174. return ret;
  175. out = (float *)frame->data[0];
  176. for (i=0; i < RA288_BLOCKS_PER_FRAME; i++) {
  177. float gain = amptable[get_bits(&gb, 3)];
  178. int cb_coef = get_bits(&gb, 6 + (i&1));
  179. decode(ractx, gain, cb_coef);
  180. memcpy(out, &ractx->sp_hist[70 + 36], RA288_BLOCK_SIZE * sizeof(*out));
  181. out += RA288_BLOCK_SIZE;
  182. if ((i & 7) == 3) {
  183. backward_filter(ractx, ractx->sp_hist, ractx->sp_rec, syn_window,
  184. ractx->sp_lpc, syn_bw_tab, 36, 40, 35, 70);
  185. backward_filter(ractx, ractx->gain_hist, ractx->gain_rec, gain_window,
  186. ractx->gain_lpc, gain_bw_tab, 10, 8, 20, 28);
  187. }
  188. }
  189. *got_frame_ptr = 1;
  190. return avctx->block_align;
  191. }
  192. AVCodec ff_ra_288_decoder = {
  193. .name = "real_288",
  194. .long_name = NULL_IF_CONFIG_SMALL("RealAudio 2.0 (28.8K)"),
  195. .type = AVMEDIA_TYPE_AUDIO,
  196. .id = AV_CODEC_ID_RA_288,
  197. .priv_data_size = sizeof(RA288Context),
  198. .init = ra288_decode_init,
  199. .decode = ra288_decode_frame,
  200. .close = ra288_decode_close,
  201. .capabilities = AV_CODEC_CAP_DR1,
  202. };