1 /******************************************************************************
2 * *
3 * Copyright (C) 2023 The Android Open Source Project
4 *
5 * Licensed under the Apache License, Version 2.0 (the "License");
6 * you may not use this file except in compliance with the License.
7 * You may obtain a copy of the License at:
8 *
9 * http://www.apache.org/licenses/LICENSE-2.0
10 *
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
16 *
17 *****************************************************************************
18 * Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore
19 */
20
21 #include <math.h>
22 #include <stdlib.h>
23 #include <string.h>
24
25 #include "ixheaac_type_def.h"
26 #include "ixheaac_constants.h"
27 #include "ixheaace_aac_constants.h"
28 #include "ixheaac_basic_ops32.h"
29 #include "ixheaac_basic_ops16.h"
30 #include "ixheaac_basic_ops40.h"
31 #include "ixheaac_basic_ops.h"
32
33 #include "ixheaace_psy_const.h"
34 #include "ixheaace_tns.h"
35 #include "ixheaace_tns_params.h"
36 #include "ixheaace_rom.h"
37 #include "ixheaace_common_rom.h"
38 #include "ixheaace_bitbuffer.h"
39
40 #include "ixheaace_block_switch.h"
41 #include "ixheaace_psy_data.h"
42 #include "ixheaace_interface.h"
43 #include "ixheaace_adjust_threshold_data.h"
44 #include "ixheaace_dynamic_bits.h"
45 #include "ixheaace_qc_data.h"
46 #include "ixheaac_error_standards.h"
47 #include "ixheaace_stereo_preproc.h"
48 #include "ixheaace_common_utils.h"
49
iaace_init_stereo_pre_processing(ixheaace_stereo_pre_pro_pstr pstr_stereo_pre_pro,WORD32 no_channels,WORD32 bit_rate,WORD32 sample_rate,FLOAT32 used_scf_ratio)50 IA_ERRORCODE iaace_init_stereo_pre_processing(ixheaace_stereo_pre_pro_pstr pstr_stereo_pre_pro,
51 WORD32 no_channels, WORD32 bit_rate, WORD32 sample_rate,
52 FLOAT32 used_scf_ratio) {
53 FLOAT32 bpf = bit_rate * 1024.0f / sample_rate;
54 FLOAT32 tmp;
55
56 memset(pstr_stereo_pre_pro, 0, sizeof(ixheaace_stereo_pre_pro_struct));
57
58 if (no_channels == 2) {
59 (pstr_stereo_pre_pro)->stereo_attenuation_flag = 1;
60
61 (pstr_stereo_pre_pro)->norm_pe_fac = 230.0f * used_scf_ratio / bpf;
62 (pstr_stereo_pre_pro)->impact_factor =
63 MIN(1, 400000.0f / (FLOAT32)(bit_rate - sample_rate * sample_rate / 72000.0f));
64 (pstr_stereo_pre_pro)->inc_stereo_attenuation = 22050.0f / sample_rate * 400.0f / bpf;
65 (pstr_stereo_pre_pro)->dec_stereo_attenuation = 22050.0f / sample_rate * 200.0f / bpf;
66
67 (pstr_stereo_pre_pro)->const_attenuation = 0.0f;
68 (pstr_stereo_pre_pro)->stereo_attenuation_max = 12.0f;
69
70 /* energy ratio thresholds (dB) */
71 (pstr_stereo_pre_pro)->sm_min = 0.0f;
72 (pstr_stereo_pre_pro)->sm_max = 15.0f;
73 (pstr_stereo_pre_pro)->lr_min = 10.0f;
74 (pstr_stereo_pre_pro)->lr_max = 30.0f;
75
76 /* pe thresholds */
77 (pstr_stereo_pre_pro)->pe_crit = 1200.0f;
78 (pstr_stereo_pre_pro)->pe_min = 700.0f;
79 (pstr_stereo_pre_pro)->pe_impact_max = 100.0f;
80
81 /* init start values */
82 (pstr_stereo_pre_pro)->average_freq_energy_l = 0.0f;
83 (pstr_stereo_pre_pro)->average_freq_energy_r = 0.0f;
84 (pstr_stereo_pre_pro)->average_freq_energy_s = 0.0f;
85 (pstr_stereo_pre_pro)->average_freq_energy_m = 0.0f;
86 (pstr_stereo_pre_pro)->smoothed_pe_sum_sum = 7000.0f; /* typical start value */
87 (pstr_stereo_pre_pro)->avg_s_to_m = -10.0f; /* typical start value */
88 (pstr_stereo_pre_pro)->last_l_to_r = 0.0f;
89 (pstr_stereo_pre_pro)->last_nrg_lr = 0.0f;
90
91 tmp = 1.0f - (bpf / 2600.0f);
92
93 tmp = MAX(tmp, 0.0f);
94
95 (pstr_stereo_pre_pro)->stereo_attenuation =
96 tmp * (pstr_stereo_pre_pro)->stereo_attenuation_max;
97 }
98
99 return IA_NO_ERROR;
100 }
101
iaace_apply_stereo_preproc(ixheaace_stereo_pre_pro_pstr pstr_stereo_pre_pro,WORD32 num_channels,ixheaace_element_info * pstr_elem_info,FLOAT32 * ptr_time_data,WORD32 granule_len)102 VOID iaace_apply_stereo_preproc(ixheaace_stereo_pre_pro_pstr pstr_stereo_pre_pro,
103 WORD32 num_channels, ixheaace_element_info *pstr_elem_info,
104 FLOAT32 *ptr_time_data, WORD32 granule_len) {
105 FLOAT32 sm_ratio, s_to_m;
106 FLOAT32 lr_ratio, l_to_r, delta_l_to_r, delta_nrg;
107 FLOAT32 en_impact, pe_impact, pe_norm;
108 FLOAT32 att, att_aimed;
109 FLOAT32 max_inc, max_dec, swift_factor;
110 FLOAT32 delta = 0.1f;
111 FLOAT32 fac = pstr_stereo_pre_pro->stereo_attenuation_fac;
112 FLOAT32 m_part, upper, div;
113 FLOAT32 l_fac, r_fac;
114 FLOAT32 L, R;
115 WORD32 i;
116
117 if (!pstr_stereo_pre_pro->stereo_attenuation_flag) {
118 return;
119 }
120
121 m_part = 2.0f * pstr_stereo_pre_pro->average_freq_energy_m * (1.0f - fac * fac);
122
123 upper = pstr_stereo_pre_pro->average_freq_energy_l * (1.0f + fac) +
124 pstr_stereo_pre_pro->average_freq_energy_r * (1.0f - fac) - m_part;
125 div = pstr_stereo_pre_pro->average_freq_energy_r * (1.0f + fac) +
126 pstr_stereo_pre_pro->average_freq_energy_l * (1.0f - fac) - m_part;
127
128 if ((div == 0.0f) || (upper == 0.0f)) {
129 l_to_r = pstr_stereo_pre_pro->lr_max;
130 } else {
131 lr_ratio = (FLOAT32)fabs(upper / div);
132 l_to_r = (FLOAT32)fabs(10.0f * log10(lr_ratio));
133 }
134
135 /* Calculate delta energy to previous frame */
136 delta_nrg = (pstr_stereo_pre_pro->average_freq_energy_l +
137 pstr_stereo_pre_pro->average_freq_energy_r + 1.0f) /
138 (pstr_stereo_pre_pro->last_nrg_lr + 1.0f);
139
140 delta_nrg = (FLOAT32)(fabs(10.0f * log10(delta_nrg)));
141
142 /* Smooth S/M over time */
143 sm_ratio = (pstr_stereo_pre_pro->average_freq_energy_s + 1.0f) /
144 (pstr_stereo_pre_pro->average_freq_energy_m + 1.0f);
145
146 s_to_m = (FLOAT32)(10.0f * log10(sm_ratio));
147
148 pstr_stereo_pre_pro->avg_s_to_m =
149 delta * s_to_m + (1 - delta) * pstr_stereo_pre_pro->avg_s_to_m;
150
151 en_impact = 1.0f;
152
153 if (pstr_stereo_pre_pro->avg_s_to_m > pstr_stereo_pre_pro->sm_min) {
154 if (pstr_stereo_pre_pro->avg_s_to_m > pstr_stereo_pre_pro->sm_max) {
155 en_impact = 0.0f;
156 } else {
157 en_impact = (pstr_stereo_pre_pro->sm_max - pstr_stereo_pre_pro->avg_s_to_m) /
158 (pstr_stereo_pre_pro->sm_max - pstr_stereo_pre_pro->sm_min);
159 }
160 }
161
162 if (l_to_r > pstr_stereo_pre_pro->lr_min) {
163 if (l_to_r > pstr_stereo_pre_pro->lr_max) {
164 en_impact = 0.0f;
165 } else {
166 en_impact *= (pstr_stereo_pre_pro->lr_max - l_to_r) /
167 (pstr_stereo_pre_pro->lr_max - pstr_stereo_pre_pro->lr_min);
168 }
169 }
170
171 pe_impact = 0.0f;
172
173 pe_norm = pstr_stereo_pre_pro->smoothed_pe_sum_sum * pstr_stereo_pre_pro->norm_pe_fac;
174
175 if (pe_norm > pstr_stereo_pre_pro->pe_min) {
176 pe_impact = ((pe_norm - pstr_stereo_pre_pro->pe_min) /
177 (pstr_stereo_pre_pro->pe_crit - pstr_stereo_pre_pro->pe_min));
178 }
179
180 pe_impact = MIN(pe_impact, pstr_stereo_pre_pro->pe_impact_max);
181
182 att_aimed = MIN((en_impact * pe_impact * pstr_stereo_pre_pro->impact_factor),
183 pstr_stereo_pre_pro->stereo_attenuation_max);
184
185 /* Only accept changes if they are large enough */
186 if ((fabs(att_aimed - pstr_stereo_pre_pro->stereo_attenuation) < 1.0f) && (att_aimed != 0.0f)) {
187 att_aimed = pstr_stereo_pre_pro->stereo_attenuation;
188 }
189
190 att = att_aimed;
191
192 swift_factor = (6.0f + pstr_stereo_pre_pro->stereo_attenuation) / (10.0f + l_to_r) *
193 MAX(1.0f, 0.2f * delta_nrg);
194
195 delta_l_to_r = MAX(3.0f, l_to_r - pstr_stereo_pre_pro->last_l_to_r);
196
197 max_dec = MIN((delta_l_to_r * delta_l_to_r / 9.0f * swift_factor), 5.0f);
198
199 max_dec *= pstr_stereo_pre_pro->dec_stereo_attenuation;
200
201 max_dec = MIN(max_dec, pstr_stereo_pre_pro->stereo_attenuation * 0.8f);
202
203 delta_l_to_r = MAX(3.0f, pstr_stereo_pre_pro->last_l_to_r - l_to_r);
204
205 max_inc = MIN((delta_l_to_r * delta_l_to_r / 9.0f * swift_factor), 5.0f);
206
207 max_inc *= pstr_stereo_pre_pro->inc_stereo_attenuation;
208
209 att = MIN(att, pstr_stereo_pre_pro->stereo_attenuation + max_inc);
210
211 att = MAX(att, pstr_stereo_pre_pro->stereo_attenuation - max_dec);
212
213 pstr_stereo_pre_pro->stereo_attenuation = (pstr_stereo_pre_pro->const_attenuation == 0)
214 ? att
215 : pstr_stereo_pre_pro->const_attenuation;
216
217 /* Perform attenuation of side channel */
218 pstr_stereo_pre_pro->stereo_attenuation_fac =
219 (FLOAT32)pow(10.0f, 0.05f * (-pstr_stereo_pre_pro->stereo_attenuation));
220
221 l_fac = 0.5f * (1.0f + pstr_stereo_pre_pro->stereo_attenuation_fac);
222 r_fac = 0.5f * (1.0f - pstr_stereo_pre_pro->stereo_attenuation_fac);
223
224 for (i = 0; i < granule_len; i++) {
225 L = l_fac * ptr_time_data[num_channels * i + pstr_elem_info->channel_index[0]] +
226 r_fac * ptr_time_data[num_channels * i + pstr_elem_info->channel_index[1]];
227 R = r_fac * ptr_time_data[num_channels * i + pstr_elem_info->channel_index[0]] +
228 l_fac * ptr_time_data[num_channels * i + pstr_elem_info->channel_index[1]];
229
230 ptr_time_data[num_channels * i + pstr_elem_info->channel_index[0]] = L;
231 ptr_time_data[num_channels * i + pstr_elem_info->channel_index[1]] = R;
232 }
233
234 pstr_stereo_pre_pro->last_l_to_r = l_to_r;
235
236 pstr_stereo_pre_pro->last_nrg_lr =
237 pstr_stereo_pre_pro->average_freq_energy_l + pstr_stereo_pre_pro->average_freq_energy_r;
238 }
239
iaace_update_stereo_pre_process(ixheaace_psy_out_channel ** pstr_psy_out,ixheaace_qc_out_element * pstr_qc_out,ixheaace_stereo_pre_pro_pstr pstr_stereo_pre_pro,FLOAT32 weight_pe_fac)240 VOID iaace_update_stereo_pre_process(ixheaace_psy_out_channel **pstr_psy_out,
241 ixheaace_qc_out_element *pstr_qc_out,
242 ixheaace_stereo_pre_pro_pstr pstr_stereo_pre_pro,
243 FLOAT32 weight_pe_fac) {
244 if (pstr_stereo_pre_pro->stereo_attenuation_flag) {
245 FLOAT32 delta = 0.1f;
246
247 pstr_stereo_pre_pro->average_freq_energy_l = pstr_psy_out[0]->sfb_sum_lr_energy;
248 pstr_stereo_pre_pro->average_freq_energy_r = pstr_psy_out[1]->sfb_sum_lr_energy;
249 pstr_stereo_pre_pro->average_freq_energy_m = pstr_psy_out[0]->sfb_sum_ms_energy;
250 pstr_stereo_pre_pro->average_freq_energy_s = pstr_psy_out[1]->sfb_sum_ms_energy;
251
252 pstr_stereo_pre_pro->smoothed_pe_sum_sum =
253 delta * pstr_qc_out->pe * weight_pe_fac +
254 (1 - delta) * pstr_stereo_pre_pro->smoothed_pe_sum_sum;
255 }
256 }
257