acelp_pitch_delay.h 8.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253
  1. /*
  2. * gain code, gain pitch and pitch delay decoding
  3. *
  4. * Copyright (c) 2008 Vladimir Voroshilov
  5. *
  6. * This file is part of FFmpeg.
  7. *
  8. * FFmpeg is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU Lesser General Public
  10. * License as published by the Free Software Foundation; either
  11. * version 2.1 of the License, or (at your option) any later version.
  12. *
  13. * FFmpeg is distributed in the hope that it will be useful,
  14. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  16. * Lesser General Public License for more details.
  17. *
  18. * You should have received a copy of the GNU Lesser General Public
  19. * License along with FFmpeg; if not, write to the Free Software
  20. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21. */
  22. #ifndef AVCODEC_ACELP_PITCH_DELAY_H
  23. #define AVCODEC_ACELP_PITCH_DELAY_H
  24. #include <stdint.h>
  25. #include "audiodsp.h"
  26. #define PITCH_DELAY_MIN 20
  27. #define PITCH_DELAY_MAX 143
  28. /**
  29. * @brief Decode pitch delay of the first subframe encoded by 8 bits with 1/3
  30. * resolution.
  31. * @param ac_index adaptive codebook index (8 bits)
  32. *
  33. * @return pitch delay in 1/3 units
  34. *
  35. * Pitch delay is coded:
  36. * with 1/3 resolution, 19 < pitch_delay < 85
  37. * integers only, 85 <= pitch_delay <= 143
  38. */
  39. int ff_acelp_decode_8bit_to_1st_delay3(int ac_index);
  40. /**
  41. * @brief Decode pitch delay of the second subframe encoded by 5 or 6 bits
  42. * with 1/3 precision.
  43. * @param ac_index adaptive codebook index (5 or 6 bits)
  44. * @param pitch_delay_min lower bound (integer) of pitch delay interval
  45. * for second subframe
  46. *
  47. * @return pitch delay in 1/3 units
  48. *
  49. * Pitch delay is coded:
  50. * with 1/3 resolution, -6 < pitch_delay - int(prev_pitch_delay) < 5
  51. *
  52. * @remark The routine is used in G.729 @@8k, AMR @@10.2k, AMR @@7.95k,
  53. * AMR @@7.4k for the second subframe.
  54. */
  55. int ff_acelp_decode_5_6_bit_to_2nd_delay3(
  56. int ac_index,
  57. int pitch_delay_min);
  58. /**
  59. * @brief Decode pitch delay with 1/3 precision.
  60. * @param ac_index adaptive codebook index (4 bits)
  61. * @param pitch_delay_min lower bound (integer) of pitch delay interval for
  62. * second subframe
  63. *
  64. * @return pitch delay in 1/3 units
  65. *
  66. * Pitch delay is coded:
  67. * integers only, -6 < pitch_delay - int(prev_pitch_delay) <= -2
  68. * with 1/3 resolution, -2 < pitch_delay - int(prev_pitch_delay) < 1
  69. * integers only, 1 <= pitch_delay - int(prev_pitch_delay) < 5
  70. *
  71. * @remark The routine is used in G.729 @@6.4k, AMR @@6.7k, AMR @@5.9k,
  72. * AMR @@5.15k, AMR @@4.75k for the second subframe.
  73. */
  74. int ff_acelp_decode_4bit_to_2nd_delay3(
  75. int ac_index,
  76. int pitch_delay_min);
  77. /**
  78. * @brief Decode pitch delay of the first subframe encoded by 9 bits
  79. * with 1/6 precision.
  80. * @param ac_index adaptive codebook index (9 bits)
  81. *
  82. * @return pitch delay in 1/6 units
  83. *
  84. * Pitch delay is coded:
  85. * with 1/6 resolution, 17 < pitch_delay < 95
  86. * integers only, 95 <= pitch_delay <= 143
  87. *
  88. * @remark The routine is used in AMR @@12.2k for the first and third subframes.
  89. */
  90. int ff_acelp_decode_9bit_to_1st_delay6(int ac_index);
  91. /**
  92. * @brief Decode pitch delay of the second subframe encoded by 6 bits
  93. * with 1/6 precision.
  94. * @param ac_index adaptive codebook index (6 bits)
  95. * @param pitch_delay_min lower bound (integer) of pitch delay interval for
  96. * second subframe
  97. *
  98. * @return pitch delay in 1/6 units
  99. *
  100. * Pitch delay is coded:
  101. * with 1/6 resolution, -6 < pitch_delay - int(prev_pitch_delay) < 5
  102. *
  103. * @remark The routine is used in AMR @@12.2k for the second and fourth subframes.
  104. */
  105. int ff_acelp_decode_6bit_to_2nd_delay6(
  106. int ac_index,
  107. int pitch_delay_min);
  108. /**
  109. * @brief Update past quantized energies
  110. * @param[in,out] quant_energy past quantized energies (5.10)
  111. * @param gain_corr_factor gain correction factor
  112. * @param log2_ma_pred_order log2() of MA prediction order
  113. * @param erasure frame erasure flag
  114. *
  115. * If frame erasure flag is not equal to zero, memory is updated with
  116. * averaged energy, attenuated by 4dB:
  117. * max(avg(quant_energy[i])-4, -14), i=0,ma_pred_order
  118. *
  119. * In normal mode memory is updated with
  120. * Er - Ep = 20 * log10(gain_corr_factor)
  121. *
  122. * @remark The routine is used in G.729 and AMR (all modes).
  123. */
  124. void ff_acelp_update_past_gain(
  125. int16_t* quant_energy,
  126. int gain_corr_factor,
  127. int log2_ma_pred_order,
  128. int erasure);
  129. /**
  130. * @brief Decode the adaptive codebook gain and add
  131. * correction (4.1.5 and 3.9.1 of G.729).
  132. * @param adsp initialized audio DSP context
  133. * @param gain_corr_factor gain correction factor (2.13)
  134. * @param fc_v fixed-codebook vector (2.13)
  135. * @param mr_energy mean innovation energy and fixed-point correction (7.13)
  136. * @param[in,out] quant_energy past quantized energies (5.10)
  137. * @param subframe_size length of subframe
  138. *
  139. * @return quantized fixed-codebook gain (14.1)
  140. *
  141. * The routine implements equations 69, 66 and 71 of the G.729 specification (3.9.1)
  142. *
  143. * Em - mean innovation energy (dB, constant, depends on decoding algorithm)
  144. * Ep - mean-removed predicted energy (dB)
  145. * Er - mean-removed innovation energy (dB)
  146. * Ei - mean energy of the fixed-codebook contribution (dB)
  147. * N - subframe_size
  148. * M - MA (Moving Average) prediction order
  149. * gc - fixed-codebook gain
  150. * gc_p - predicted fixed-codebook gain
  151. *
  152. * Fixed codebook gain is computed using predicted gain gc_p and
  153. * correction factor gain_corr_factor as shown below:
  154. *
  155. * gc = gc_p * gain_corr_factor
  156. *
  157. * The predicted fixed codebook gain gc_p is found by predicting
  158. * the energy of the fixed-codebook contribution from the energy
  159. * of previous fixed-codebook contributions.
  160. *
  161. * mean = 1/N * sum(i,0,N){ fc_v[i] * fc_v[i] }
  162. *
  163. * Ei = 10log(mean)
  164. *
  165. * Er = 10log(1/N * gc^2 * mean) - Em = 20log(gc) + Ei - Em
  166. *
  167. * Replacing Er with Ep and gc with gc_p we will receive:
  168. *
  169. * Ep = 10log(1/N * gc_p^2 * mean) - Em = 20log(gc_p) + Ei - Em
  170. *
  171. * and from above:
  172. *
  173. * gc_p = 10^((Ep - Ei + Em) / 20)
  174. *
  175. * Ep is predicted using past energies and prediction coefficients:
  176. *
  177. * Ep = sum(i,0,M){ ma_prediction_coeff[i] * quant_energy[i] }
  178. *
  179. * gc_p in fixed-point arithmetic is calculated as following:
  180. *
  181. * mean = 1/N * sum(i,0,N){ (fc_v[i] / 2^13) * (fc_v[i] / 2^13) } =
  182. * = 1/N * sum(i,0,N) { fc_v[i] * fc_v[i] } / 2^26
  183. *
  184. * Ei = 10log(mean) = -10log(N) - 10log(2^26) +
  185. * + 10log(sum(i,0,N) { fc_v[i] * fc_v[i] })
  186. *
  187. * Ep - Ei + Em = Ep + Em + 10log(N) + 10log(2^26) -
  188. * - 10log(sum(i,0,N) { fc_v[i] * fc_v[i] }) =
  189. * = Ep + mr_energy - 10log(sum(i,0,N) { fc_v[i] * fc_v[i] })
  190. *
  191. * gc_p = 10 ^ ((Ep - Ei + Em) / 20) =
  192. * = 2 ^ (3.3219 * (Ep - Ei + Em) / 20) = 2 ^ (0.166 * (Ep - Ei + Em))
  193. *
  194. * where
  195. *
  196. * mr_energy = Em + 10log(N) + 10log(2^26)
  197. *
  198. * @remark The routine is used in G.729 and AMR (all modes).
  199. */
  200. int16_t ff_acelp_decode_gain_code(
  201. AudioDSPContext *adsp,
  202. int gain_corr_factor,
  203. const int16_t* fc_v,
  204. int mr_energy,
  205. const int16_t* quant_energy,
  206. const int16_t* ma_prediction_coeff,
  207. int subframe_size,
  208. int max_pred_order);
  209. /**
  210. * Calculate fixed gain (part of section 6.1.3 of AMR spec)
  211. *
  212. * @param fixed_gain_factor gain correction factor
  213. * @param fixed_mean_energy mean decoded algebraic codebook vector energy
  214. * @param prediction_error vector of the quantified predictor errors of
  215. * the four previous subframes. It is updated by this function.
  216. * @param energy_mean desired mean innovation energy
  217. * @param pred_table table of four moving average coefficients
  218. */
  219. float ff_amr_set_fixed_gain(float fixed_gain_factor, float fixed_mean_energy,
  220. float *prediction_error, float energy_mean,
  221. const float *pred_table);
  222. /**
  223. * Decode the adaptive codebook index to the integer and fractional parts
  224. * of the pitch lag for one subframe at 1/3 fractional precision.
  225. *
  226. * The choice of pitch lag is described in 3GPP TS 26.090 section 5.6.1.
  227. *
  228. * @param lag_int integer part of pitch lag of the current subframe
  229. * @param lag_frac fractional part of pitch lag of the current subframe
  230. * @param pitch_index parsed adaptive codebook (pitch) index
  231. * @param prev_lag_int integer part of pitch lag for the previous subframe
  232. * @param subframe current subframe number
  233. * @param third_as_first treat the third frame the same way as the first
  234. */
  235. void ff_decode_pitch_lag(int *lag_int, int *lag_frac, int pitch_index,
  236. const int prev_lag_int, const int subframe,
  237. int third_as_first, int resolution);
  238. #endif /* AVCODEC_ACELP_PITCH_DELAY_H */