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 */

23 #ifndef AVCODEC_ACELP_PITCH_DELAY_H

24 #define AVCODEC_ACELP_PITCH_DELAY_H

26 #include <stdint.h>

29 #define PITCH_DELAY_MIN 20

30 #define PITCH_DELAY_MAX 143

32 /**

33 * \brief Decode pitch delay of the first subframe encoded by 8 bits with 1/3

34 * resolution.

35 * \param ac_index adaptive codebook index (8 bits)

36 *

37 * \return pitch delay in 1/3 units

38 *

39 * Pitch delay is coded:

40 * with 1/3 resolution, 19 < pitch_delay < 85

41 * integers only, 85 <= pitch_delay <= 143

42 */

45 /**

46 * \brief Decode pitch delay of the second subframe encoded by 5 or 6 bits

47 * with 1/3 precision.

48 * \param ac_index adaptive codebook index (5 or 6 bits)

49 * \param pitch_delay_min lower bound (integer) of pitch delay interval

50 * for second subframe

51 *

52 * \return pitch delay in 1/3 units

53 *

54 * Pitch delay is coded:

55 * with 1/3 resolution, -6 < pitch_delay - int(prev_pitch_delay) < 5

56 *

57 * \remark The routine is used in G.729 @@8k, AMR @@10.2k, AMR @@7.95k,

58 * AMR @@7.4k for the second subframe.

59 */

64 /**

65 * \brief Decode pitch delay with 1/3 precision.

66 * \param ac_index adaptive codebook index (4 bits)

67 * \param pitch_delay_min lower bound (integer) of pitch delay interval for

68 * second subframe

69 *

70 * \return pitch delay in 1/3 units

71 *

72 * Pitch delay is coded:

73 * integers only, -6 < pitch_delay - int(prev_pitch_delay) <= -2

74 * with 1/3 resolution, -2 < pitch_delay - int(prev_pitch_delay) < 1

75 * integers only, 1 <= pitch_delay - int(prev_pitch_delay) < 5

76 *

77 * \remark The routine is used in G.729 @@6.4k, AMR @@6.7k, AMR @@5.9k,

78 * AMR @@5.15k, AMR @@4.75k for the second subframe.

79 */

84 /**

85 * \brief Decode pitch delay of the first subframe encoded by 9 bits

86 * with 1/6 precision.

87 * \param ac_index adaptive codebook index (9 bits)

88 * \param pitch_delay_min lower bound (integer) of pitch delay interval for

89 * second subframe

90 *

91 * \return pitch delay in 1/6 units

92 *

93 * Pitch delay is coded:

94 * with 1/6 resolution, 17 < pitch_delay < 95

95 * integers only, 95 <= pitch_delay <= 143

96 *

97 * \remark The routine is used in AMR @@12.2k for the first and third subframes.

98 */

101 /**

102 * \brief Decode pitch delay of the second subframe encoded by 6 bits

103 * with 1/6 precision.

104 * \param ac_index adaptive codebook index (6 bits)

105 * \param pitch_delay_min lower bound (integer) of pitch delay interval for

106 * second subframe

107 *

108 * \return pitch delay in 1/6 units

109 *

110 * Pitch delay is coded:

111 * with 1/6 resolution, -6 < pitch_delay - int(prev_pitch_delay) < 5

112 *

113 * \remark The routine is used in AMR @@12.2k for the second and fourth subframes.

114 */

119 /**

120 * \brief Update past quantized energies

121 * \param quant_energy [in/out] past quantized energies (5.10)

122 * \param gain_corr_factor gain correction factor

123 * \param log2_ma_pred_order log2() of MA prediction order

124 * \param erasure frame erasure flag

125 *

126 * If frame erasure flag is not equal to zero, memory is updated with

127 * averaged energy, attenuated by 4dB:

128 * max(avg(quant_energy[i])-4, -14), i=0,ma_pred_order

129 *

130 * In normal mode memory is updated with

131 * Er - Ep = 20 * log10(gain_corr_factor)

132 *

133 * \remark The routine is used in G.729 and AMR (all modes).

134 */

141 /**

142 * \brief Decode the adaptive codebook gain and add

143 * correction (4.1.5 and 3.9.1 of G.729).

144 * \param dsp initialized dsputil context

145 * \param gain_corr_factor gain correction factor (2.13)

146 * \param fc_v fixed-codebook vector (2.13)

147 * \param mr_energy mean innovation energy and fixed-point correction (7.13)

148 * \param quant_energy [in/out] past quantized energies (5.10)

149 * \param subframe_size length of subframe

150 * \param ma_pred_order MA prediction order

151 *

152 * \return quantized fixed-codebook gain (14.1)

153 *

154 * The routine implements equations 69, 66 and 71 of the G.729 specification (3.9.1)

155 *

156 * Em - mean innovation energy (dB, constant, depends on decoding algorithm)

157 * Ep - mean-removed predicted energy (dB)

158 * Er - mean-removed innovation energy (dB)

159 * Ei - mean energy of the fixed-codebook contribution (dB)

160 * N - subframe_size

161 * M - MA (Moving Average) prediction order

162 * gc - fixed-codebook gain

163 * gc_p - predicted fixed-codebook gain

164 *

165 * Fixed codebook gain is computed using predicted gain gc_p and

166 * correction factor gain_corr_factor as shown below:

167 *

168 * gc = gc_p * gain_corr_factor

169 *

170 * The predicted fixed codebook gain gc_p is found by predicting

171 * the energy of the fixed-codebook contribution from the energy

172 * of previous fixed-codebook contributions.

173 *

174 * mean = 1/N * sum(i,0,N){ fc_v[i] * fc_v[i] }

175 *

176 * Ei = 10log(mean)

177 *

178 * Er = 10log(1/N * gc^2 * mean) - Em = 20log(gc) + Ei - Em

179 *

180 * Replacing Er with Ep and gc with gc_p we will receive:

181 *

182 * Ep = 10log(1/N * gc_p^2 * mean) - Em = 20log(gc_p) + Ei - Em

183 *

184 * and from above:

185 *

186 * gc_p = 10^((Ep - Ei + Em) / 20)

187 *

188 * Ep is predicted using past energies and prediction coefficients:

189 *

190 * Ep = sum(i,0,M){ ma_prediction_coeff[i] * quant_energy[i] }

191 *

192 * gc_p in fixed-point arithmetic is calculated as following:

193 *

194 * mean = 1/N * sum(i,0,N){ (fc_v[i] / 2^13) * (fc_v[i] / 2^13) } =

195 * = 1/N * sum(i,0,N) { fc_v[i] * fc_v[i] } / 2^26

196 *

197 * Ei = 10log(mean) = -10log(N) - 10log(2^26) +

198 * + 10log(sum(i,0,N) { fc_v[i] * fc_v[i] })

199 *

200 * Ep - Ei + Em = Ep + Em + 10log(N) + 10log(2^26) -

201 * - 10log(sum(i,0,N) { fc_v[i] * fc_v[i] }) =

202 * = Ep + mr_energy - 10log(sum(i,0,N) { fc_v[i] * fc_v[i] })

203 *

204 * gc_p = 10 ^ ((Ep - Ei + Em) / 20) =

205 * = 2 ^ (3.3219 * (Ep - Ei + Em) / 20) = 2 ^ (0.166 * (Ep - Ei + Em))

206 *

207 * where

208 *

209 * mr_energy = Em + 10log(N) + 10log(2^26)

210 *

211 * \remark The routine is used in G.729 and AMR (all modes).

212 */

223 /**

224 * Calculate fixed gain (part of section 6.1.3 of AMR spec)

225 *

226 * @param fixed_gain_factor gain correction factor

227 * @param fixed_energy decoded algebraic codebook vector energy

228 * @param prediction_error vector of the quantified predictor errors of

229 * the four previous subframes. It is updated by this function.

230 * @param energy_mean desired mean innovation energy

231 * @param pred_table table of four moving average coefficients

232 */

238 /**

239 * Decode the adaptive codebook index to the integer and fractional parts

240 * of the pitch lag for one subframe at 1/3 fractional precision.

241 *

242 * The choice of pitch lag is described in 3GPP TS 26.090 section 5.6.1.

243 *

244 * @param lag_int integer part of pitch lag of the current subframe

245 * @param lag_frac fractional part of pitch lag of the current subframe

246 * @param pitch_index parsed adaptive codebook (pitch) index

247 * @param prev_lag_int integer part of pitch lag for the previous subframe

248 * @param subframe current subframe number

249 * @param third_as_first treat the third frame the same way as the first

250 */