e5410bba7f57487aebefeb94e399343762f45ccc
[libav.git] / libavcodec / acelp_pitch_delay.h
1 /*
2 * gain code, gain pitch and pitch delay decoding
3 *
4 * Copyright (c) 2008 Vladimir Voroshilov
5 *
6 * This file is part of Libav.
7 *
8 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 #ifndef AVCODEC_ACELP_PITCH_DELAY_H
24 #define AVCODEC_ACELP_PITCH_DELAY_H
25
26 #include <stdint.h>
27 #include "dsputil.h"
28
29 #define PITCH_DELAY_MIN 20
30 #define PITCH_DELAY_MAX 143
31
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 */
43 int ff_acelp_decode_8bit_to_1st_delay3(int ac_index);
44
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 */
60 int ff_acelp_decode_5_6_bit_to_2nd_delay3(
61 int ac_index,
62 int pitch_delay_min);
63
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 */
80 int ff_acelp_decode_4bit_to_2nd_delay3(
81 int ac_index,
82 int pitch_delay_min);
83
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 *
89 * @return pitch delay in 1/6 units
90 *
91 * Pitch delay is coded:
92 * with 1/6 resolution, 17 < pitch_delay < 95
93 * integers only, 95 <= pitch_delay <= 143
94 *
95 * @remark The routine is used in AMR @@12.2k for the first and third subframes.
96 */
97 int ff_acelp_decode_9bit_to_1st_delay6(int ac_index);
98
99 /**
100 * @brief Decode pitch delay of the second subframe encoded by 6 bits
101 * with 1/6 precision.
102 * @param ac_index adaptive codebook index (6 bits)
103 * @param pitch_delay_min lower bound (integer) of pitch delay interval for
104 * second subframe
105 *
106 * @return pitch delay in 1/6 units
107 *
108 * Pitch delay is coded:
109 * with 1/6 resolution, -6 < pitch_delay - int(prev_pitch_delay) < 5
110 *
111 * @remark The routine is used in AMR @@12.2k for the second and fourth subframes.
112 */
113 int ff_acelp_decode_6bit_to_2nd_delay6(
114 int ac_index,
115 int pitch_delay_min);
116
117 /**
118 * @brief Update past quantized energies
119 * @param[in,out] quant_energy past quantized energies (5.10)
120 * @param gain_corr_factor gain correction factor
121 * @param log2_ma_pred_order log2() of MA prediction order
122 * @param erasure frame erasure flag
123 *
124 * If frame erasure flag is not equal to zero, memory is updated with
125 * averaged energy, attenuated by 4dB:
126 * max(avg(quant_energy[i])-4, -14), i=0,ma_pred_order
127 *
128 * In normal mode memory is updated with
129 * Er - Ep = 20 * log10(gain_corr_factor)
130 *
131 * @remark The routine is used in G.729 and AMR (all modes).
132 */
133 void ff_acelp_update_past_gain(
134 int16_t* quant_energy,
135 int gain_corr_factor,
136 int log2_ma_pred_order,
137 int erasure);
138
139 /**
140 * @brief Decode the adaptive codebook gain and add
141 * correction (4.1.5 and 3.9.1 of G.729).
142 * @param dsp initialized dsputil context
143 * @param gain_corr_factor gain correction factor (2.13)
144 * @param fc_v fixed-codebook vector (2.13)
145 * @param mr_energy mean innovation energy and fixed-point correction (7.13)
146 * @param[in,out] quant_energy past quantized energies (5.10)
147 * @param subframe_size length of subframe
148 *
149 * @return quantized fixed-codebook gain (14.1)
150 *
151 * The routine implements equations 69, 66 and 71 of the G.729 specification (3.9.1)
152 *
153 * Em - mean innovation energy (dB, constant, depends on decoding algorithm)
154 * Ep - mean-removed predicted energy (dB)
155 * Er - mean-removed innovation energy (dB)
156 * Ei - mean energy of the fixed-codebook contribution (dB)
157 * N - subframe_size
158 * M - MA (Moving Average) prediction order
159 * gc - fixed-codebook gain
160 * gc_p - predicted fixed-codebook gain
161 *
162 * Fixed codebook gain is computed using predicted gain gc_p and
163 * correction factor gain_corr_factor as shown below:
164 *
165 * gc = gc_p * gain_corr_factor
166 *
167 * The predicted fixed codebook gain gc_p is found by predicting
168 * the energy of the fixed-codebook contribution from the energy
169 * of previous fixed-codebook contributions.
170 *
171 * mean = 1/N * sum(i,0,N){ fc_v[i] * fc_v[i] }
172 *
173 * Ei = 10log(mean)
174 *
175 * Er = 10log(1/N * gc^2 * mean) - Em = 20log(gc) + Ei - Em
176 *
177 * Replacing Er with Ep and gc with gc_p we will receive:
178 *
179 * Ep = 10log(1/N * gc_p^2 * mean) - Em = 20log(gc_p) + Ei - Em
180 *
181 * and from above:
182 *
183 * gc_p = 10^((Ep - Ei + Em) / 20)
184 *
185 * Ep is predicted using past energies and prediction coefficients:
186 *
187 * Ep = sum(i,0,M){ ma_prediction_coeff[i] * quant_energy[i] }
188 *
189 * gc_p in fixed-point arithmetic is calculated as following:
190 *
191 * mean = 1/N * sum(i,0,N){ (fc_v[i] / 2^13) * (fc_v[i] / 2^13) } =
192 * = 1/N * sum(i,0,N) { fc_v[i] * fc_v[i] } / 2^26
193 *
194 * Ei = 10log(mean) = -10log(N) - 10log(2^26) +
195 * + 10log(sum(i,0,N) { fc_v[i] * fc_v[i] })
196 *
197 * Ep - Ei + Em = Ep + Em + 10log(N) + 10log(2^26) -
198 * - 10log(sum(i,0,N) { fc_v[i] * fc_v[i] }) =
199 * = Ep + mr_energy - 10log(sum(i,0,N) { fc_v[i] * fc_v[i] })
200 *
201 * gc_p = 10 ^ ((Ep - Ei + Em) / 20) =
202 * = 2 ^ (3.3219 * (Ep - Ei + Em) / 20) = 2 ^ (0.166 * (Ep - Ei + Em))
203 *
204 * where
205 *
206 * mr_energy = Em + 10log(N) + 10log(2^26)
207 *
208 * @remark The routine is used in G.729 and AMR (all modes).
209 */
210 int16_t ff_acelp_decode_gain_code(
211 DSPContext *dsp,
212 int gain_corr_factor,
213 const int16_t* fc_v,
214 int mr_energy,
215 const int16_t* quant_energy,
216 const int16_t* ma_prediction_coeff,
217 int subframe_size,
218 int max_pred_order);
219
220 /**
221 * Calculate fixed gain (part of section 6.1.3 of AMR spec)
222 *
223 * @param fixed_gain_factor gain correction factor
224 * @param fixed_mean_energy mean decoded algebraic codebook vector energy
225 * @param prediction_error vector of the quantified predictor errors of
226 * the four previous subframes. It is updated by this function.
227 * @param energy_mean desired mean innovation energy
228 * @param pred_table table of four moving average coefficients
229 */
230 float ff_amr_set_fixed_gain(float fixed_gain_factor, float fixed_mean_energy,
231 float *prediction_error, float energy_mean,
232 const float *pred_table);
233
234
235 /**
236 * Decode the adaptive codebook index to the integer and fractional parts
237 * of the pitch lag for one subframe at 1/3 fractional precision.
238 *
239 * The choice of pitch lag is described in 3GPP TS 26.090 section 5.6.1.
240 *
241 * @param lag_int integer part of pitch lag of the current subframe
242 * @param lag_frac fractional part of pitch lag of the current subframe
243 * @param pitch_index parsed adaptive codebook (pitch) index
244 * @param prev_lag_int integer part of pitch lag for the previous subframe
245 * @param subframe current subframe number
246 * @param third_as_first treat the third frame the same way as the first
247 */
248 void ff_decode_pitch_lag(int *lag_int, int *lag_frac, int pitch_index,
249 const int prev_lag_int, const int subframe,
250 int third_as_first, int resolution);
251
252 #endif /* AVCODEC_ACELP_PITCH_DELAY_H */