/* * Copyright (c) 2012 The WebRTC project authors. All Rights Reserved. * * Use of this source code is governed by a BSD-style license * that can be found in the LICENSE file in the root of the source * tree. An additional intellectual property rights grant can be found * in the file PATENTS. All contributing project authors may * be found in the AUTHORS file in the root of the source tree. */ #include "webrtc/modules/audio_processing/ns/include/noise_suppression_x.h" #include #include #include #include #include "webrtc/common_audio/signal_processing/include/real_fft.h" #include "webrtc/modules/audio_processing/ns/nsx_core.h" #include "webrtc/system_wrappers/include/cpu_features_wrapper.h" #if (defined WEBRTC_DETECT_NEON || defined WEBRTC_HAS_NEON) /* Tables are defined in ARM assembly files. */ extern const int16_t WebRtcNsx_kLogTable[9]; extern const int16_t WebRtcNsx_kCounterDiv[201]; extern const int16_t WebRtcNsx_kLogTableFrac[256]; #else static const int16_t WebRtcNsx_kLogTable[9] = { 0, 177, 355, 532, 710, 887, 1065, 1242, 1420 }; static const int16_t WebRtcNsx_kCounterDiv[201] = { 32767, 16384, 10923, 8192, 6554, 5461, 4681, 4096, 3641, 3277, 2979, 2731, 2521, 2341, 2185, 2048, 1928, 1820, 1725, 1638, 1560, 1489, 1425, 1365, 1311, 1260, 1214, 1170, 1130, 1092, 1057, 1024, 993, 964, 936, 910, 886, 862, 840, 819, 799, 780, 762, 745, 728, 712, 697, 683, 669, 655, 643, 630, 618, 607, 596, 585, 575, 565, 555, 546, 537, 529, 520, 512, 504, 496, 489, 482, 475, 468, 462, 455, 449, 443, 437, 431, 426, 420, 415, 410, 405, 400, 395, 390, 386, 381, 377, 372, 368, 364, 360, 356, 352, 349, 345, 341, 338, 334, 331, 328, 324, 321, 318, 315, 312, 309, 306, 303, 301, 298, 295, 293, 290, 287, 285, 282, 280, 278, 275, 273, 271, 269, 266, 264, 262, 260, 258, 256, 254, 252, 250, 248, 246, 245, 243, 241, 239, 237, 236, 234, 232, 231, 229, 228, 226, 224, 223, 221, 220, 218, 217, 216, 214, 213, 211, 210, 209, 207, 206, 205, 204, 202, 201, 200, 199, 197, 196, 195, 194, 193, 192, 191, 189, 188, 187, 186, 185, 184, 183, 182, 181, 180, 179, 178, 177, 176, 175, 174, 173, 172, 172, 171, 170, 169, 168, 167, 166, 165, 165, 164, 163 }; static const int16_t WebRtcNsx_kLogTableFrac[256] = { 0, 1, 3, 4, 6, 7, 9, 10, 11, 13, 14, 16, 17, 18, 20, 21, 22, 24, 25, 26, 28, 29, 30, 32, 33, 34, 36, 37, 38, 40, 41, 42, 44, 45, 46, 47, 49, 50, 51, 52, 54, 55, 56, 57, 59, 60, 61, 62, 63, 65, 66, 67, 68, 69, 71, 72, 73, 74, 75, 77, 78, 79, 80, 81, 82, 84, 85, 86, 87, 88, 89, 90, 92, 93, 94, 95, 96, 97, 98, 99, 100, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 178, 179, 180, 181, 182, 183, 184, 185, 185, 186, 187, 188, 189, 190, 191, 192, 192, 193, 194, 195, 196, 197, 198, 198, 199, 200, 201, 202, 203, 203, 204, 205, 206, 207, 208, 208, 209, 210, 211, 212, 212, 213, 214, 215, 216, 216, 217, 218, 219, 220, 220, 221, 222, 223, 224, 224, 225, 226, 227, 228, 228, 229, 230, 231, 231, 232, 233, 234, 234, 235, 236, 237, 238, 238, 239, 240, 241, 241, 242, 243, 244, 244, 245, 246, 247, 247, 248, 249, 249, 250, 251, 252, 252, 253, 254, 255, 255 }; #endif // WEBRTC_DETECT_NEON || WEBRTC_HAS_NEON // Skip first frequency bins during estimation. (0 <= value < 64) static const size_t kStartBand = 5; // hybrib Hanning & flat window static const int16_t kBlocks80w128x[128] = { 0, 536, 1072, 1606, 2139, 2669, 3196, 3720, 4240, 4756, 5266, 5771, 6270, 6762, 7246, 7723, 8192, 8652, 9102, 9543, 9974, 10394, 10803, 11200, 11585, 11958, 12318, 12665, 12998, 13318, 13623, 13913, 14189, 14449, 14694, 14924, 15137, 15334, 15515, 15679, 15826, 15956, 16069, 16165, 16244, 16305, 16349, 16375, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16375, 16349, 16305, 16244, 16165, 16069, 15956, 15826, 15679, 15515, 15334, 15137, 14924, 14694, 14449, 14189, 13913, 13623, 13318, 12998, 12665, 12318, 11958, 11585, 11200, 10803, 10394, 9974, 9543, 9102, 8652, 8192, 7723, 7246, 6762, 6270, 5771, 5266, 4756, 4240, 3720, 3196, 2669, 2139, 1606, 1072, 536 }; // hybrib Hanning & flat window static const int16_t kBlocks160w256x[256] = { 0, 268, 536, 804, 1072, 1339, 1606, 1872, 2139, 2404, 2669, 2933, 3196, 3459, 3720, 3981, 4240, 4499, 4756, 5012, 5266, 5520, 5771, 6021, 6270, 6517, 6762, 7005, 7246, 7486, 7723, 7959, 8192, 8423, 8652, 8878, 9102, 9324, 9543, 9760, 9974, 10185, 10394, 10600, 10803, 11003, 11200, 11394, 11585, 11773, 11958, 12140, 12318, 12493, 12665, 12833, 12998, 13160, 13318, 13472, 13623, 13770, 13913, 14053, 14189, 14321, 14449, 14574, 14694, 14811, 14924, 15032, 15137, 15237, 15334, 15426, 15515, 15599, 15679, 15754, 15826, 15893, 15956, 16015, 16069, 16119, 16165, 16207, 16244, 16277, 16305, 16329, 16349, 16364, 16375, 16382, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16382, 16375, 16364, 16349, 16329, 16305, 16277, 16244, 16207, 16165, 16119, 16069, 16015, 15956, 15893, 15826, 15754, 15679, 15599, 15515, 15426, 15334, 15237, 15137, 15032, 14924, 14811, 14694, 14574, 14449, 14321, 14189, 14053, 13913, 13770, 13623, 13472, 13318, 13160, 12998, 12833, 12665, 12493, 12318, 12140, 11958, 11773, 11585, 11394, 11200, 11003, 10803, 10600, 10394, 10185, 9974, 9760, 9543, 9324, 9102, 8878, 8652, 8423, 8192, 7959, 7723, 7486, 7246, 7005, 6762, 6517, 6270, 6021, 5771, 5520, 5266, 5012, 4756, 4499, 4240, 3981, 3720, 3459, 3196, 2933, 2669, 2404, 2139, 1872, 1606, 1339, 1072, 804, 536, 268 }; // Gain factor1 table: Input value in Q8 and output value in Q13 // original floating point code // if (gain > blim) { // factor1 = 1.0 + 1.3 * (gain - blim); // if (gain * factor1 > 1.0) { // factor1 = 1.0 / gain; // } // } else { // factor1 = 1.0; // } static const int16_t kFactor1Table[257] = { 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8233, 8274, 8315, 8355, 8396, 8436, 8475, 8515, 8554, 8592, 8631, 8669, 8707, 8745, 8783, 8820, 8857, 8894, 8931, 8967, 9003, 9039, 9075, 9111, 9146, 9181, 9216, 9251, 9286, 9320, 9354, 9388, 9422, 9456, 9489, 9523, 9556, 9589, 9622, 9655, 9687, 9719, 9752, 9784, 9816, 9848, 9879, 9911, 9942, 9973, 10004, 10035, 10066, 10097, 10128, 10158, 10188, 10218, 10249, 10279, 10308, 10338, 10368, 10397, 10426, 10456, 10485, 10514, 10543, 10572, 10600, 10629, 10657, 10686, 10714, 10742, 10770, 10798, 10826, 10854, 10882, 10847, 10810, 10774, 10737, 10701, 10666, 10631, 10596, 10562, 10527, 10494, 10460, 10427, 10394, 10362, 10329, 10297, 10266, 10235, 10203, 10173, 10142, 10112, 10082, 10052, 10023, 9994, 9965, 9936, 9908, 9879, 9851, 9824, 9796, 9769, 9742, 9715, 9689, 9662, 9636, 9610, 9584, 9559, 9534, 9508, 9484, 9459, 9434, 9410, 9386, 9362, 9338, 9314, 9291, 9268, 9245, 9222, 9199, 9176, 9154, 9132, 9110, 9088, 9066, 9044, 9023, 9002, 8980, 8959, 8939, 8918, 8897, 8877, 8857, 8836, 8816, 8796, 8777, 8757, 8738, 8718, 8699, 8680, 8661, 8642, 8623, 8605, 8586, 8568, 8550, 8532, 8514, 8496, 8478, 8460, 8443, 8425, 8408, 8391, 8373, 8356, 8339, 8323, 8306, 8289, 8273, 8256, 8240, 8224, 8208, 8192 }; // For Factor2 tables // original floating point code // if (gain > blim) { // factor2 = 1.0; // } else { // factor2 = 1.0 - 0.3 * (blim - gain); // if (gain <= inst->denoiseBound) { // factor2 = 1.0 - 0.3 * (blim - inst->denoiseBound); // } // } // // Gain factor table: Input value in Q8 and output value in Q13 static const int16_t kFactor2Aggressiveness1[257] = { 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7577, 7596, 7614, 7632, 7650, 7667, 7683, 7699, 7715, 7731, 7746, 7761, 7775, 7790, 7804, 7818, 7832, 7845, 7858, 7871, 7884, 7897, 7910, 7922, 7934, 7946, 7958, 7970, 7982, 7993, 8004, 8016, 8027, 8038, 8049, 8060, 8070, 8081, 8091, 8102, 8112, 8122, 8132, 8143, 8152, 8162, 8172, 8182, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192 }; // Gain factor table: Input value in Q8 and output value in Q13 static const int16_t kFactor2Aggressiveness2[257] = { 7270, 7270, 7270, 7270, 7270, 7306, 7339, 7369, 7397, 7424, 7448, 7472, 7495, 7517, 7537, 7558, 7577, 7596, 7614, 7632, 7650, 7667, 7683, 7699, 7715, 7731, 7746, 7761, 7775, 7790, 7804, 7818, 7832, 7845, 7858, 7871, 7884, 7897, 7910, 7922, 7934, 7946, 7958, 7970, 7982, 7993, 8004, 8016, 8027, 8038, 8049, 8060, 8070, 8081, 8091, 8102, 8112, 8122, 8132, 8143, 8152, 8162, 8172, 8182, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192 }; // Gain factor table: Input value in Q8 and output value in Q13 static const int16_t kFactor2Aggressiveness3[257] = { 7184, 7184, 7184, 7229, 7270, 7306, 7339, 7369, 7397, 7424, 7448, 7472, 7495, 7517, 7537, 7558, 7577, 7596, 7614, 7632, 7650, 7667, 7683, 7699, 7715, 7731, 7746, 7761, 7775, 7790, 7804, 7818, 7832, 7845, 7858, 7871, 7884, 7897, 7910, 7922, 7934, 7946, 7958, 7970, 7982, 7993, 8004, 8016, 8027, 8038, 8049, 8060, 8070, 8081, 8091, 8102, 8112, 8122, 8132, 8143, 8152, 8162, 8172, 8182, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192 }; // sum of log2(i) from table index to inst->anaLen2 in Q5 // Note that the first table value is invalid, since log2(0) = -infinity static const int16_t kSumLogIndex[66] = { 0, 22917, 22917, 22885, 22834, 22770, 22696, 22613, 22524, 22428, 22326, 22220, 22109, 21994, 21876, 21754, 21629, 21501, 21370, 21237, 21101, 20963, 20822, 20679, 20535, 20388, 20239, 20089, 19937, 19783, 19628, 19470, 19312, 19152, 18991, 18828, 18664, 18498, 18331, 18164, 17994, 17824, 17653, 17480, 17306, 17132, 16956, 16779, 16602, 16423, 16243, 16063, 15881, 15699, 15515, 15331, 15146, 14960, 14774, 14586, 14398, 14209, 14019, 13829, 13637, 13445 }; // sum of log2(i)^2 from table index to inst->anaLen2 in Q2 // Note that the first table value is invalid, since log2(0) = -infinity static const int16_t kSumSquareLogIndex[66] = { 0, 16959, 16959, 16955, 16945, 16929, 16908, 16881, 16850, 16814, 16773, 16729, 16681, 16630, 16575, 16517, 16456, 16392, 16325, 16256, 16184, 16109, 16032, 15952, 15870, 15786, 15700, 15612, 15521, 15429, 15334, 15238, 15140, 15040, 14938, 14834, 14729, 14622, 14514, 14404, 14292, 14179, 14064, 13947, 13830, 13710, 13590, 13468, 13344, 13220, 13094, 12966, 12837, 12707, 12576, 12444, 12310, 12175, 12039, 11902, 11763, 11624, 11483, 11341, 11198, 11054 }; // log2(table index) in Q12 // Note that the first table value is invalid, since log2(0) = -infinity static const int16_t kLogIndex[129] = { 0, 0, 4096, 6492, 8192, 9511, 10588, 11499, 12288, 12984, 13607, 14170, 14684, 15157, 15595, 16003, 16384, 16742, 17080, 17400, 17703, 17991, 18266, 18529, 18780, 19021, 19253, 19476, 19691, 19898, 20099, 20292, 20480, 20662, 20838, 21010, 21176, 21338, 21496, 21649, 21799, 21945, 22087, 22226, 22362, 22495, 22625, 22752, 22876, 22998, 23117, 23234, 23349, 23462, 23572, 23680, 23787, 23892, 23994, 24095, 24195, 24292, 24388, 24483, 24576, 24668, 24758, 24847, 24934, 25021, 25106, 25189, 25272, 25354, 25434, 25513, 25592, 25669, 25745, 25820, 25895, 25968, 26041, 26112, 26183, 26253, 26322, 26390, 26458, 26525, 26591, 26656, 26721, 26784, 26848, 26910, 26972, 27033, 27094, 27154, 27213, 27272, 27330, 27388, 27445, 27502, 27558, 27613, 27668, 27722, 27776, 27830, 27883, 27935, 27988, 28039, 28090, 28141, 28191, 28241, 28291, 28340, 28388, 28437, 28484, 28532, 28579, 28626, 28672 }; // determinant of estimation matrix in Q0 corresponding to the log2 tables above // Note that the first table value is invalid, since log2(0) = -infinity static const int16_t kDeterminantEstMatrix[66] = { 0, 29814, 25574, 22640, 20351, 18469, 16873, 15491, 14277, 13199, 12233, 11362, 10571, 9851, 9192, 8587, 8030, 7515, 7038, 6596, 6186, 5804, 5448, 5115, 4805, 4514, 4242, 3988, 3749, 3524, 3314, 3116, 2930, 2755, 2590, 2435, 2289, 2152, 2022, 1900, 1785, 1677, 1575, 1478, 1388, 1302, 1221, 1145, 1073, 1005, 942, 881, 825, 771, 721, 674, 629, 587, 547, 510, 475, 442, 411, 382, 355, 330 }; // Update the noise estimation information. static void UpdateNoiseEstimate(NoiseSuppressionFixedC* inst, int offset) { int32_t tmp32no1 = 0; int32_t tmp32no2 = 0; int16_t tmp16 = 0; const int16_t kExp2Const = 11819; // Q13 size_t i = 0; tmp16 = WebRtcSpl_MaxValueW16(inst->noiseEstLogQuantile + offset, inst->magnLen); // Guarantee a Q-domain as high as possible and still fit in int16 inst->qNoise = 14 - (int) WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND( kExp2Const, tmp16, 21); for (i = 0; i < inst->magnLen; i++) { // inst->quantile[i]=exp(inst->lquantile[offset+i]); // in Q21 tmp32no2 = kExp2Const * inst->noiseEstLogQuantile[offset + i]; tmp32no1 = (0x00200000 | (tmp32no2 & 0x001FFFFF)); // 2^21 + frac tmp16 = (int16_t)(tmp32no2 >> 21); tmp16 -= 21;// shift 21 to get result in Q0 tmp16 += (int16_t) inst->qNoise; //shift to get result in Q(qNoise) if (tmp16 < 0) { tmp32no1 >>= -tmp16; } else { tmp32no1 <<= tmp16; } inst->noiseEstQuantile[i] = WebRtcSpl_SatW32ToW16(tmp32no1); } } // Noise Estimation static void NoiseEstimationC(NoiseSuppressionFixedC* inst, uint16_t* magn, uint32_t* noise, int16_t* q_noise) { int16_t lmagn[HALF_ANAL_BLOCKL], counter, countDiv; int16_t countProd, delta, zeros, frac; int16_t log2, tabind, logval, tmp16, tmp16no1, tmp16no2; const int16_t log2_const = 22713; // Q15 const int16_t width_factor = 21845; size_t i, s, offset; tabind = inst->stages - inst->normData; assert(tabind < 9); assert(tabind > -9); if (tabind < 0) { logval = -WebRtcNsx_kLogTable[-tabind]; } else { logval = WebRtcNsx_kLogTable[tabind]; } // lmagn(i)=log(magn(i))=log(2)*log2(magn(i)) // magn is in Q(-stages), and the real lmagn values are: // real_lmagn(i)=log(magn(i)*2^stages)=log(magn(i))+log(2^stages) // lmagn in Q8 for (i = 0; i < inst->magnLen; i++) { if (magn[i]) { zeros = WebRtcSpl_NormU32((uint32_t)magn[i]); frac = (int16_t)((((uint32_t)magn[i] << zeros) & 0x7FFFFFFF) >> 23); // log2(magn(i)) assert(frac < 256); log2 = (int16_t)(((31 - zeros) << 8) + WebRtcNsx_kLogTableFrac[frac]); // log2(magn(i))*log(2) lmagn[i] = (int16_t)((log2 * log2_const) >> 15); // + log(2^stages) lmagn[i] += logval; } else { lmagn[i] = logval;//0; } } // loop over simultaneous estimates for (s = 0; s < SIMULT; s++) { offset = s * inst->magnLen; // Get counter values from state counter = inst->noiseEstCounter[s]; assert(counter < 201); countDiv = WebRtcNsx_kCounterDiv[counter]; countProd = (int16_t)(counter * countDiv); // quant_est(...) for (i = 0; i < inst->magnLen; i++) { // compute delta if (inst->noiseEstDensity[offset + i] > 512) { // Get the value for delta by shifting intead of dividing. int factor = WebRtcSpl_NormW16(inst->noiseEstDensity[offset + i]); delta = (int16_t)(FACTOR_Q16 >> (14 - factor)); } else { delta = FACTOR_Q7; if (inst->blockIndex < END_STARTUP_LONG) { // Smaller step size during startup. This prevents from using // unrealistic values causing overflow. delta = FACTOR_Q7_STARTUP; } } // update log quantile estimate tmp16 = (int16_t)((delta * countDiv) >> 14); if (lmagn[i] > inst->noiseEstLogQuantile[offset + i]) { // +=QUANTILE*delta/(inst->counter[s]+1) QUANTILE=0.25, =1 in Q2 // CounterDiv=1/(inst->counter[s]+1) in Q15 tmp16 += 2; inst->noiseEstLogQuantile[offset + i] += tmp16 / 4; } else { tmp16 += 1; // *(1-QUANTILE), in Q2 QUANTILE=0.25, 1-0.25=0.75=3 in Q2 // TODO(bjornv): investigate why we need to truncate twice. tmp16no2 = (int16_t)((tmp16 / 2) * 3 / 2); inst->noiseEstLogQuantile[offset + i] -= tmp16no2; if (inst->noiseEstLogQuantile[offset + i] < logval) { // This is the smallest fixed point representation we can // have, hence we limit the output. inst->noiseEstLogQuantile[offset + i] = logval; } } // update density estimate if (WEBRTC_SPL_ABS_W16(lmagn[i] - inst->noiseEstLogQuantile[offset + i]) < WIDTH_Q8) { tmp16no1 = (int16_t)WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND( inst->noiseEstDensity[offset + i], countProd, 15); tmp16no2 = (int16_t)WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND( width_factor, countDiv, 15); inst->noiseEstDensity[offset + i] = tmp16no1 + tmp16no2; } } // end loop over magnitude spectrum if (counter >= END_STARTUP_LONG) { inst->noiseEstCounter[s] = 0; if (inst->blockIndex >= END_STARTUP_LONG) { UpdateNoiseEstimate(inst, offset); } } inst->noiseEstCounter[s]++; } // end loop over simultaneous estimates // Sequentially update the noise during startup if (inst->blockIndex < END_STARTUP_LONG) { UpdateNoiseEstimate(inst, offset); } for (i = 0; i < inst->magnLen; i++) { noise[i] = (uint32_t)(inst->noiseEstQuantile[i]); // Q(qNoise) } (*q_noise) = (int16_t)inst->qNoise; } // Filter the data in the frequency domain, and create spectrum. static void PrepareSpectrumC(NoiseSuppressionFixedC* inst, int16_t* freq_buf) { size_t i = 0, j = 0; for (i = 0; i < inst->magnLen; i++) { inst->real[i] = (int16_t)((inst->real[i] * (int16_t)(inst->noiseSupFilter[i])) >> 14); // Q(normData-stages) inst->imag[i] = (int16_t)((inst->imag[i] * (int16_t)(inst->noiseSupFilter[i])) >> 14); // Q(normData-stages) } freq_buf[0] = inst->real[0]; freq_buf[1] = -inst->imag[0]; for (i = 1, j = 2; i < inst->anaLen2; i += 1, j += 2) { freq_buf[j] = inst->real[i]; freq_buf[j + 1] = -inst->imag[i]; } freq_buf[inst->anaLen] = inst->real[inst->anaLen2]; freq_buf[inst->anaLen + 1] = -inst->imag[inst->anaLen2]; } // Denormalize the real-valued signal |in|, the output from inverse FFT. static void DenormalizeC(NoiseSuppressionFixedC* inst, int16_t* in, int factor) { size_t i = 0; int32_t tmp32 = 0; for (i = 0; i < inst->anaLen; i += 1) { tmp32 = WEBRTC_SPL_SHIFT_W32((int32_t)in[i], factor - inst->normData); inst->real[i] = WebRtcSpl_SatW32ToW16(tmp32); // Q0 } } // For the noise supression process, synthesis, read out fully processed // segment, and update synthesis buffer. static void SynthesisUpdateC(NoiseSuppressionFixedC* inst, int16_t* out_frame, int16_t gain_factor) { size_t i = 0; int16_t tmp16a = 0; int16_t tmp16b = 0; int32_t tmp32 = 0; // synthesis for (i = 0; i < inst->anaLen; i++) { tmp16a = (int16_t)WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND( inst->window[i], inst->real[i], 14); // Q0, window in Q14 tmp32 = WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND(tmp16a, gain_factor, 13); // Q0 // Down shift with rounding tmp16b = WebRtcSpl_SatW32ToW16(tmp32); // Q0 inst->synthesisBuffer[i] = WebRtcSpl_AddSatW16(inst->synthesisBuffer[i], tmp16b); // Q0 } // read out fully processed segment for (i = 0; i < inst->blockLen10ms; i++) { out_frame[i] = inst->synthesisBuffer[i]; // Q0 } // update synthesis buffer memcpy(inst->synthesisBuffer, inst->synthesisBuffer + inst->blockLen10ms, (inst->anaLen - inst->blockLen10ms) * sizeof(*inst->synthesisBuffer)); WebRtcSpl_ZerosArrayW16(inst->synthesisBuffer + inst->anaLen - inst->blockLen10ms, inst->blockLen10ms); } // Update analysis buffer for lower band, and window data before FFT. static void AnalysisUpdateC(NoiseSuppressionFixedC* inst, int16_t* out, int16_t* new_speech) { size_t i = 0; // For lower band update analysis buffer. memcpy(inst->analysisBuffer, inst->analysisBuffer + inst->blockLen10ms, (inst->anaLen - inst->blockLen10ms) * sizeof(*inst->analysisBuffer)); memcpy(inst->analysisBuffer + inst->anaLen - inst->blockLen10ms, new_speech, inst->blockLen10ms * sizeof(*inst->analysisBuffer)); // Window data before FFT. for (i = 0; i < inst->anaLen; i++) { out[i] = (int16_t)WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND( inst->window[i], inst->analysisBuffer[i], 14); // Q0 } } // Normalize the real-valued signal |in|, the input to forward FFT. static void NormalizeRealBufferC(NoiseSuppressionFixedC* inst, const int16_t* in, int16_t* out) { size_t i = 0; assert(inst->normData >= 0); for (i = 0; i < inst->anaLen; ++i) { out[i] = in[i] << inst->normData; // Q(normData) } } // Declare function pointers. NoiseEstimation WebRtcNsx_NoiseEstimation; PrepareSpectrum WebRtcNsx_PrepareSpectrum; SynthesisUpdate WebRtcNsx_SynthesisUpdate; AnalysisUpdate WebRtcNsx_AnalysisUpdate; Denormalize WebRtcNsx_Denormalize; NormalizeRealBuffer WebRtcNsx_NormalizeRealBuffer; #if (defined WEBRTC_DETECT_NEON || defined WEBRTC_HAS_NEON) // Initialize function pointers for ARM Neon platform. static void WebRtcNsx_InitNeon(void) { WebRtcNsx_NoiseEstimation = WebRtcNsx_NoiseEstimationNeon; WebRtcNsx_PrepareSpectrum = WebRtcNsx_PrepareSpectrumNeon; WebRtcNsx_SynthesisUpdate = WebRtcNsx_SynthesisUpdateNeon; WebRtcNsx_AnalysisUpdate = WebRtcNsx_AnalysisUpdateNeon; } #endif #if defined(MIPS32_LE) // Initialize function pointers for MIPS platform. static void WebRtcNsx_InitMips(void) { WebRtcNsx_PrepareSpectrum = WebRtcNsx_PrepareSpectrum_mips; WebRtcNsx_SynthesisUpdate = WebRtcNsx_SynthesisUpdate_mips; WebRtcNsx_AnalysisUpdate = WebRtcNsx_AnalysisUpdate_mips; WebRtcNsx_NormalizeRealBuffer = WebRtcNsx_NormalizeRealBuffer_mips; #if defined(MIPS_DSP_R1_LE) WebRtcNsx_Denormalize = WebRtcNsx_Denormalize_mips; #endif } #endif void WebRtcNsx_CalcParametricNoiseEstimate(NoiseSuppressionFixedC* inst, int16_t pink_noise_exp_avg, int32_t pink_noise_num_avg, int freq_index, uint32_t* noise_estimate, uint32_t* noise_estimate_avg) { int32_t tmp32no1 = 0; int32_t tmp32no2 = 0; int16_t int_part = 0; int16_t frac_part = 0; // Use pink noise estimate // noise_estimate = 2^(pinkNoiseNumerator + pinkNoiseExp * log2(j)) assert(freq_index >= 0); assert(freq_index < 129); tmp32no2 = (pink_noise_exp_avg * kLogIndex[freq_index]) >> 15; // Q11 tmp32no1 = pink_noise_num_avg - tmp32no2; // Q11 // Calculate output: 2^tmp32no1 // Output in Q(minNorm-stages) tmp32no1 += (inst->minNorm - inst->stages) << 11; if (tmp32no1 > 0) { int_part = (int16_t)(tmp32no1 >> 11); frac_part = (int16_t)(tmp32no1 & 0x000007ff); // Q11 // Piecewise linear approximation of 'b' in // 2^(int_part+frac_part) = 2^int_part * (1 + b) // 'b' is given in Q11 and below stored in frac_part. if (frac_part >> 10) { // Upper fractional part tmp32no2 = (2048 - frac_part) * 1244; // Q21 tmp32no2 = 2048 - (tmp32no2 >> 10); } else { // Lower fractional part tmp32no2 = (frac_part * 804) >> 10; } // Shift fractional part to Q(minNorm-stages) tmp32no2 = WEBRTC_SPL_SHIFT_W32(tmp32no2, int_part - 11); *noise_estimate_avg = (1 << int_part) + (uint32_t)tmp32no2; // Scale up to initMagnEst, which is not block averaged *noise_estimate = (*noise_estimate_avg) * (uint32_t)(inst->blockIndex + 1); } } // Initialize state int32_t WebRtcNsx_InitCore(NoiseSuppressionFixedC* inst, uint32_t fs) { int i; //check for valid pointer if (inst == NULL) { return -1; } // // Initialization of struct if (fs == 8000 || fs == 16000 || fs == 32000 || fs == 48000) { inst->fs = fs; } else { return -1; } if (fs == 8000) { inst->blockLen10ms = 80; inst->anaLen = 128; inst->stages = 7; inst->window = kBlocks80w128x; inst->thresholdLogLrt = 131072; //default threshold for LRT feature inst->maxLrt = 0x0040000; inst->minLrt = 52429; } else { inst->blockLen10ms = 160; inst->anaLen = 256; inst->stages = 8; inst->window = kBlocks160w256x; inst->thresholdLogLrt = 212644; //default threshold for LRT feature inst->maxLrt = 0x0080000; inst->minLrt = 104858; } inst->anaLen2 = inst->anaLen / 2; inst->magnLen = inst->anaLen2 + 1; if (inst->real_fft != NULL) { WebRtcSpl_FreeRealFFT(inst->real_fft); } inst->real_fft = WebRtcSpl_CreateRealFFT(inst->stages); if (inst->real_fft == NULL) { return -1; } WebRtcSpl_ZerosArrayW16(inst->analysisBuffer, ANAL_BLOCKL_MAX); WebRtcSpl_ZerosArrayW16(inst->synthesisBuffer, ANAL_BLOCKL_MAX); // for HB processing WebRtcSpl_ZerosArrayW16(inst->dataBufHBFX[0], NUM_HIGH_BANDS_MAX * ANAL_BLOCKL_MAX); // for quantile noise estimation WebRtcSpl_ZerosArrayW16(inst->noiseEstQuantile, HALF_ANAL_BLOCKL); for (i = 0; i < SIMULT * HALF_ANAL_BLOCKL; i++) { inst->noiseEstLogQuantile[i] = 2048; // Q8 inst->noiseEstDensity[i] = 153; // Q9 } for (i = 0; i < SIMULT; i++) { inst->noiseEstCounter[i] = (int16_t)(END_STARTUP_LONG * (i + 1)) / SIMULT; } // Initialize suppression filter with ones WebRtcSpl_MemSetW16((int16_t*)inst->noiseSupFilter, 16384, HALF_ANAL_BLOCKL); // Set the aggressiveness: default inst->aggrMode = 0; //initialize variables for new method inst->priorNonSpeechProb = 8192; // Q14(0.5) prior probability for speech/noise for (i = 0; i < HALF_ANAL_BLOCKL; i++) { inst->prevMagnU16[i] = 0; inst->prevNoiseU32[i] = 0; //previous noise-spectrum inst->logLrtTimeAvgW32[i] = 0; //smooth LR ratio inst->avgMagnPause[i] = 0; //conservative noise spectrum estimate inst->initMagnEst[i] = 0; //initial average magnitude spectrum } //feature quantities inst->thresholdSpecDiff = 50; //threshold for difference feature: determined on-line inst->thresholdSpecFlat = 20480; //threshold for flatness: determined on-line inst->featureLogLrt = inst->thresholdLogLrt; //average LRT factor (= threshold) inst->featureSpecFlat = inst->thresholdSpecFlat; //spectral flatness (= threshold) inst->featureSpecDiff = inst->thresholdSpecDiff; //spectral difference (= threshold) inst->weightLogLrt = 6; //default weighting par for LRT feature inst->weightSpecFlat = 0; //default weighting par for spectral flatness feature inst->weightSpecDiff = 0; //default weighting par for spectral difference feature inst->curAvgMagnEnergy = 0; //window time-average of input magnitude spectrum inst->timeAvgMagnEnergy = 0; //normalization for spectral difference inst->timeAvgMagnEnergyTmp = 0; //normalization for spectral difference //histogram quantities: used to estimate/update thresholds for features WebRtcSpl_ZerosArrayW16(inst->histLrt, HIST_PAR_EST); WebRtcSpl_ZerosArrayW16(inst->histSpecDiff, HIST_PAR_EST); WebRtcSpl_ZerosArrayW16(inst->histSpecFlat, HIST_PAR_EST); inst->blockIndex = -1; //frame counter //inst->modelUpdate = 500; //window for update inst->modelUpdate = (1 << STAT_UPDATES); //window for update inst->cntThresUpdate = 0; //counter feature thresholds updates inst->sumMagn = 0; inst->magnEnergy = 0; inst->prevQMagn = 0; inst->qNoise = 0; inst->prevQNoise = 0; inst->energyIn = 0; inst->scaleEnergyIn = 0; inst->whiteNoiseLevel = 0; inst->pinkNoiseNumerator = 0; inst->pinkNoiseExp = 0; inst->minNorm = 15; // Start with full scale inst->zeroInputSignal = 0; //default mode WebRtcNsx_set_policy_core(inst, 0); #ifdef NS_FILEDEBUG inst->infile = fopen("indebug.pcm", "wb"); inst->outfile = fopen("outdebug.pcm", "wb"); inst->file1 = fopen("file1.pcm", "wb"); inst->file2 = fopen("file2.pcm", "wb"); inst->file3 = fopen("file3.pcm", "wb"); inst->file4 = fopen("file4.pcm", "wb"); inst->file5 = fopen("file5.pcm", "wb"); #endif // Initialize function pointers. WebRtcNsx_NoiseEstimation = NoiseEstimationC; WebRtcNsx_PrepareSpectrum = PrepareSpectrumC; WebRtcNsx_SynthesisUpdate = SynthesisUpdateC; WebRtcNsx_AnalysisUpdate = AnalysisUpdateC; WebRtcNsx_Denormalize = DenormalizeC; WebRtcNsx_NormalizeRealBuffer = NormalizeRealBufferC; #ifdef WEBRTC_DETECT_NEON uint64_t features = WebRtc_GetCPUFeaturesARM(); if ((features & kCPUFeatureNEON) != 0) { WebRtcNsx_InitNeon(); } #elif defined(WEBRTC_HAS_NEON) WebRtcNsx_InitNeon(); #endif #if defined(MIPS32_LE) WebRtcNsx_InitMips(); #endif inst->initFlag = 1; return 0; } int WebRtcNsx_set_policy_core(NoiseSuppressionFixedC* inst, int mode) { // allow for modes:0,1,2,3 if (mode < 0 || mode > 3) { return -1; } inst->aggrMode = mode; if (mode == 0) { inst->overdrive = 256; // Q8(1.0) inst->denoiseBound = 8192; // Q14(0.5) inst->gainMap = 0; // No gain compensation } else if (mode == 1) { inst->overdrive = 256; // Q8(1.0) inst->denoiseBound = 4096; // Q14(0.25) inst->factor2Table = kFactor2Aggressiveness1; inst->gainMap = 1; } else if (mode == 2) { inst->overdrive = 282; // ~= Q8(1.1) inst->denoiseBound = 2048; // Q14(0.125) inst->factor2Table = kFactor2Aggressiveness2; inst->gainMap = 1; } else if (mode == 3) { inst->overdrive = 320; // Q8(1.25) inst->denoiseBound = 1475; // ~= Q14(0.09) inst->factor2Table = kFactor2Aggressiveness3; inst->gainMap = 1; } return 0; } // Extract thresholds for feature parameters // histograms are computed over some window_size (given by window_pars) // thresholds and weights are extracted every window // flag 0 means update histogram only, flag 1 means compute the thresholds/weights // threshold and weights are returned in: inst->priorModelPars void WebRtcNsx_FeatureParameterExtraction(NoiseSuppressionFixedC* inst, int flag) { uint32_t tmpU32; uint32_t histIndex; uint32_t posPeak1SpecFlatFX, posPeak2SpecFlatFX; uint32_t posPeak1SpecDiffFX, posPeak2SpecDiffFX; int32_t tmp32; int32_t fluctLrtFX, thresFluctLrtFX; int32_t avgHistLrtFX, avgSquareHistLrtFX, avgHistLrtComplFX; int16_t j; int16_t numHistLrt; int i; int useFeatureSpecFlat, useFeatureSpecDiff, featureSum; int maxPeak1, maxPeak2; int weightPeak1SpecFlat, weightPeak2SpecFlat; int weightPeak1SpecDiff, weightPeak2SpecDiff; //update histograms if (!flag) { // LRT // Type casting to UWord32 is safe since negative values will not be wrapped to larger // values than HIST_PAR_EST histIndex = (uint32_t)(inst->featureLogLrt); if (histIndex < HIST_PAR_EST) { inst->histLrt[histIndex]++; } // Spectral flatness // (inst->featureSpecFlat*20)>>10 = (inst->featureSpecFlat*5)>>8 histIndex = (inst->featureSpecFlat * 5) >> 8; if (histIndex < HIST_PAR_EST) { inst->histSpecFlat[histIndex]++; } // Spectral difference histIndex = HIST_PAR_EST; if (inst->timeAvgMagnEnergy > 0) { // Guard against division by zero // If timeAvgMagnEnergy == 0 we have no normalizing statistics and // therefore can't update the histogram histIndex = ((inst->featureSpecDiff * 5) >> inst->stages) / inst->timeAvgMagnEnergy; } if (histIndex < HIST_PAR_EST) { inst->histSpecDiff[histIndex]++; } } // extract parameters for speech/noise probability if (flag) { useFeatureSpecDiff = 1; //for LRT feature: // compute the average over inst->featureExtractionParams.rangeAvgHistLrt avgHistLrtFX = 0; avgSquareHistLrtFX = 0; numHistLrt = 0; for (i = 0; i < BIN_SIZE_LRT; i++) { j = (2 * i + 1); tmp32 = inst->histLrt[i] * j; avgHistLrtFX += tmp32; numHistLrt += inst->histLrt[i]; avgSquareHistLrtFX += tmp32 * j; } avgHistLrtComplFX = avgHistLrtFX; for (; i < HIST_PAR_EST; i++) { j = (2 * i + 1); tmp32 = inst->histLrt[i] * j; avgHistLrtComplFX += tmp32; avgSquareHistLrtFX += tmp32 * j; } fluctLrtFX = avgSquareHistLrtFX * numHistLrt - avgHistLrtFX * avgHistLrtComplFX; thresFluctLrtFX = THRES_FLUCT_LRT * numHistLrt; // get threshold for LRT feature: tmpU32 = (FACTOR_1_LRT_DIFF * (uint32_t)avgHistLrtFX); if ((fluctLrtFX < thresFluctLrtFX) || (numHistLrt == 0) || (tmpU32 > (uint32_t)(100 * numHistLrt))) { //very low fluctuation, so likely noise inst->thresholdLogLrt = inst->maxLrt; } else { tmp32 = (int32_t)((tmpU32 << (9 + inst->stages)) / numHistLrt / 25); // check if value is within min/max range inst->thresholdLogLrt = WEBRTC_SPL_SAT(inst->maxLrt, tmp32, inst->minLrt); } if (fluctLrtFX < thresFluctLrtFX) { // Do not use difference feature if fluctuation of LRT feature is very low: // most likely just noise state useFeatureSpecDiff = 0; } // for spectral flatness and spectral difference: compute the main peaks of histogram maxPeak1 = 0; maxPeak2 = 0; posPeak1SpecFlatFX = 0; posPeak2SpecFlatFX = 0; weightPeak1SpecFlat = 0; weightPeak2SpecFlat = 0; // peaks for flatness for (i = 0; i < HIST_PAR_EST; i++) { if (inst->histSpecFlat[i] > maxPeak1) { // Found new "first" peak maxPeak2 = maxPeak1; weightPeak2SpecFlat = weightPeak1SpecFlat; posPeak2SpecFlatFX = posPeak1SpecFlatFX; maxPeak1 = inst->histSpecFlat[i]; weightPeak1SpecFlat = inst->histSpecFlat[i]; posPeak1SpecFlatFX = (uint32_t)(2 * i + 1); } else if (inst->histSpecFlat[i] > maxPeak2) { // Found new "second" peak maxPeak2 = inst->histSpecFlat[i]; weightPeak2SpecFlat = inst->histSpecFlat[i]; posPeak2SpecFlatFX = (uint32_t)(2 * i + 1); } } // for spectral flatness feature useFeatureSpecFlat = 1; // merge the two peaks if they are close if ((posPeak1SpecFlatFX - posPeak2SpecFlatFX < LIM_PEAK_SPACE_FLAT_DIFF) && (weightPeak2SpecFlat * LIM_PEAK_WEIGHT_FLAT_DIFF > weightPeak1SpecFlat)) { weightPeak1SpecFlat += weightPeak2SpecFlat; posPeak1SpecFlatFX = (posPeak1SpecFlatFX + posPeak2SpecFlatFX) >> 1; } //reject if weight of peaks is not large enough, or peak value too small if (weightPeak1SpecFlat < THRES_WEIGHT_FLAT_DIFF || posPeak1SpecFlatFX < THRES_PEAK_FLAT) { useFeatureSpecFlat = 0; } else { // if selected, get the threshold // compute the threshold and check if value is within min/max range inst->thresholdSpecFlat = WEBRTC_SPL_SAT(MAX_FLAT_Q10, FACTOR_2_FLAT_Q10 * posPeak1SpecFlatFX, MIN_FLAT_Q10); //Q10 } // done with flatness feature if (useFeatureSpecDiff) { //compute two peaks for spectral difference maxPeak1 = 0; maxPeak2 = 0; posPeak1SpecDiffFX = 0; posPeak2SpecDiffFX = 0; weightPeak1SpecDiff = 0; weightPeak2SpecDiff = 0; // peaks for spectral difference for (i = 0; i < HIST_PAR_EST; i++) { if (inst->histSpecDiff[i] > maxPeak1) { // Found new "first" peak maxPeak2 = maxPeak1; weightPeak2SpecDiff = weightPeak1SpecDiff; posPeak2SpecDiffFX = posPeak1SpecDiffFX; maxPeak1 = inst->histSpecDiff[i]; weightPeak1SpecDiff = inst->histSpecDiff[i]; posPeak1SpecDiffFX = (uint32_t)(2 * i + 1); } else if (inst->histSpecDiff[i] > maxPeak2) { // Found new "second" peak maxPeak2 = inst->histSpecDiff[i]; weightPeak2SpecDiff = inst->histSpecDiff[i]; posPeak2SpecDiffFX = (uint32_t)(2 * i + 1); } } // merge the two peaks if they are close if ((posPeak1SpecDiffFX - posPeak2SpecDiffFX < LIM_PEAK_SPACE_FLAT_DIFF) && (weightPeak2SpecDiff * LIM_PEAK_WEIGHT_FLAT_DIFF > weightPeak1SpecDiff)) { weightPeak1SpecDiff += weightPeak2SpecDiff; posPeak1SpecDiffFX = (posPeak1SpecDiffFX + posPeak2SpecDiffFX) >> 1; } // get the threshold value and check if value is within min/max range inst->thresholdSpecDiff = WEBRTC_SPL_SAT(MAX_DIFF, FACTOR_1_LRT_DIFF * posPeak1SpecDiffFX, MIN_DIFF); //5x bigger //reject if weight of peaks is not large enough if (weightPeak1SpecDiff < THRES_WEIGHT_FLAT_DIFF) { useFeatureSpecDiff = 0; } // done with spectral difference feature } // select the weights between the features // inst->priorModelPars[4] is weight for LRT: always selected featureSum = 6 / (1 + useFeatureSpecFlat + useFeatureSpecDiff); inst->weightLogLrt = featureSum; inst->weightSpecFlat = useFeatureSpecFlat * featureSum; inst->weightSpecDiff = useFeatureSpecDiff * featureSum; // set histograms to zero for next update WebRtcSpl_ZerosArrayW16(inst->histLrt, HIST_PAR_EST); WebRtcSpl_ZerosArrayW16(inst->histSpecDiff, HIST_PAR_EST); WebRtcSpl_ZerosArrayW16(inst->histSpecFlat, HIST_PAR_EST); } // end of flag == 1 } // Compute spectral flatness on input spectrum // magn is the magnitude spectrum // spectral flatness is returned in inst->featureSpecFlat void WebRtcNsx_ComputeSpectralFlatness(NoiseSuppressionFixedC* inst, uint16_t* magn) { uint32_t tmpU32; uint32_t avgSpectralFlatnessNum, avgSpectralFlatnessDen; int32_t tmp32; int32_t currentSpectralFlatness, logCurSpectralFlatness; int16_t zeros, frac, intPart; size_t i; // for flatness avgSpectralFlatnessNum = 0; avgSpectralFlatnessDen = inst->sumMagn - (uint32_t)magn[0]; // Q(normData-stages) // compute log of ratio of the geometric to arithmetic mean: check for log(0) case // flatness = exp( sum(log(magn[i]))/N - log(sum(magn[i])/N) ) // = exp( sum(log(magn[i]))/N ) * N / sum(magn[i]) // = 2^( sum(log2(magn[i]))/N - (log2(sum(magn[i])) - log2(N)) ) [This is used] for (i = 1; i < inst->magnLen; i++) { // First bin is excluded from spectrum measures. Number of bins is now a power of 2 if (magn[i]) { zeros = WebRtcSpl_NormU32((uint32_t)magn[i]); frac = (int16_t)(((uint32_t)((uint32_t)(magn[i]) << zeros) & 0x7FFFFFFF) >> 23); // log2(magn(i)) assert(frac < 256); tmpU32 = (uint32_t)(((31 - zeros) << 8) + WebRtcNsx_kLogTableFrac[frac]); // Q8 avgSpectralFlatnessNum += tmpU32; // Q8 } else { //if at least one frequency component is zero, treat separately tmpU32 = WEBRTC_SPL_UMUL_32_16(inst->featureSpecFlat, SPECT_FLAT_TAVG_Q14); // Q24 inst->featureSpecFlat -= tmpU32 >> 14; // Q10 return; } } //ratio and inverse log: check for case of log(0) zeros = WebRtcSpl_NormU32(avgSpectralFlatnessDen); frac = (int16_t)(((avgSpectralFlatnessDen << zeros) & 0x7FFFFFFF) >> 23); // log2(avgSpectralFlatnessDen) assert(frac < 256); tmp32 = (int32_t)(((31 - zeros) << 8) + WebRtcNsx_kLogTableFrac[frac]); // Q8 logCurSpectralFlatness = (int32_t)avgSpectralFlatnessNum; logCurSpectralFlatness += ((int32_t)(inst->stages - 1) << (inst->stages + 7)); // Q(8+stages-1) logCurSpectralFlatness -= (tmp32 << (inst->stages - 1)); logCurSpectralFlatness <<= (10 - inst->stages); // Q17 tmp32 = (int32_t)(0x00020000 | (WEBRTC_SPL_ABS_W32(logCurSpectralFlatness) & 0x0001FFFF)); //Q17 intPart = 7 - (logCurSpectralFlatness >> 17); // Add 7 for output in Q10. if (intPart > 0) { currentSpectralFlatness = tmp32 >> intPart; } else { currentSpectralFlatness = tmp32 << -intPart; } //time average update of spectral flatness feature tmp32 = currentSpectralFlatness - (int32_t)inst->featureSpecFlat; // Q10 tmp32 *= SPECT_FLAT_TAVG_Q14; // Q24 inst->featureSpecFlat += tmp32 >> 14; // Q10 // done with flatness feature } // Compute the difference measure between input spectrum and a template/learned noise spectrum // magn_tmp is the input spectrum // the reference/template spectrum is inst->magn_avg_pause[i] // returns (normalized) spectral difference in inst->featureSpecDiff void WebRtcNsx_ComputeSpectralDifference(NoiseSuppressionFixedC* inst, uint16_t* magnIn) { // This is to be calculated: // avgDiffNormMagn = var(magnIn) - cov(magnIn, magnAvgPause)^2 / var(magnAvgPause) uint32_t tmpU32no1, tmpU32no2; uint32_t varMagnUFX, varPauseUFX, avgDiffNormMagnUFX; int32_t tmp32no1, tmp32no2; int32_t avgPauseFX, avgMagnFX, covMagnPauseFX; int32_t maxPause, minPause; int16_t tmp16no1; size_t i; int norm32, nShifts; avgPauseFX = 0; maxPause = 0; minPause = inst->avgMagnPause[0]; // Q(prevQMagn) // compute average quantities for (i = 0; i < inst->magnLen; i++) { // Compute mean of magn_pause avgPauseFX += inst->avgMagnPause[i]; // in Q(prevQMagn) maxPause = WEBRTC_SPL_MAX(maxPause, inst->avgMagnPause[i]); minPause = WEBRTC_SPL_MIN(minPause, inst->avgMagnPause[i]); } // normalize by replacing div of "inst->magnLen" with "inst->stages-1" shifts avgPauseFX >>= inst->stages - 1; avgMagnFX = inst->sumMagn >> (inst->stages - 1); // Largest possible deviation in magnPause for (co)var calculations tmp32no1 = WEBRTC_SPL_MAX(maxPause - avgPauseFX, avgPauseFX - minPause); // Get number of shifts to make sure we don't get wrap around in varPause nShifts = WEBRTC_SPL_MAX(0, 10 + inst->stages - WebRtcSpl_NormW32(tmp32no1)); varMagnUFX = 0; varPauseUFX = 0; covMagnPauseFX = 0; for (i = 0; i < inst->magnLen; i++) { // Compute var and cov of magn and magn_pause tmp16no1 = (int16_t)((int32_t)magnIn[i] - avgMagnFX); tmp32no2 = inst->avgMagnPause[i] - avgPauseFX; varMagnUFX += (uint32_t)(tmp16no1 * tmp16no1); // Q(2*qMagn) tmp32no1 = tmp32no2 * tmp16no1; // Q(prevQMagn+qMagn) covMagnPauseFX += tmp32no1; // Q(prevQMagn+qMagn) tmp32no1 = tmp32no2 >> nShifts; // Q(prevQMagn-minPause). varPauseUFX += tmp32no1 * tmp32no1; // Q(2*(prevQMagn-minPause)) } //update of average magnitude spectrum: Q(-2*stages) and averaging replaced by shifts inst->curAvgMagnEnergy += inst->magnEnergy >> (2 * inst->normData + inst->stages - 1); avgDiffNormMagnUFX = varMagnUFX; // Q(2*qMagn) if ((varPauseUFX) && (covMagnPauseFX)) { tmpU32no1 = (uint32_t)WEBRTC_SPL_ABS_W32(covMagnPauseFX); // Q(prevQMagn+qMagn) norm32 = WebRtcSpl_NormU32(tmpU32no1) - 16; if (norm32 > 0) { tmpU32no1 <<= norm32; // Q(prevQMagn+qMagn+norm32) } else { tmpU32no1 >>= -norm32; // Q(prevQMagn+qMagn+norm32) } tmpU32no2 = WEBRTC_SPL_UMUL(tmpU32no1, tmpU32no1); // Q(2*(prevQMagn+qMagn-norm32)) nShifts += norm32; nShifts <<= 1; if (nShifts < 0) { varPauseUFX >>= (-nShifts); // Q(2*(qMagn+norm32+minPause)) nShifts = 0; } if (varPauseUFX > 0) { // Q(2*(qMagn+norm32-16+minPause)) tmpU32no1 = tmpU32no2 / varPauseUFX; tmpU32no1 >>= nShifts; // Q(2*qMagn) avgDiffNormMagnUFX -= WEBRTC_SPL_MIN(avgDiffNormMagnUFX, tmpU32no1); } else { avgDiffNormMagnUFX = 0; } } //normalize and compute time average update of difference feature tmpU32no1 = avgDiffNormMagnUFX >> (2 * inst->normData); if (inst->featureSpecDiff > tmpU32no1) { tmpU32no2 = WEBRTC_SPL_UMUL_32_16(inst->featureSpecDiff - tmpU32no1, SPECT_DIFF_TAVG_Q8); // Q(8-2*stages) inst->featureSpecDiff -= tmpU32no2 >> 8; // Q(-2*stages) } else { tmpU32no2 = WEBRTC_SPL_UMUL_32_16(tmpU32no1 - inst->featureSpecDiff, SPECT_DIFF_TAVG_Q8); // Q(8-2*stages) inst->featureSpecDiff += tmpU32no2 >> 8; // Q(-2*stages) } } // Transform input (speechFrame) to frequency domain magnitude (magnU16) void WebRtcNsx_DataAnalysis(NoiseSuppressionFixedC* inst, short* speechFrame, uint16_t* magnU16) { uint32_t tmpU32no1; int32_t tmp_1_w32 = 0; int32_t tmp_2_w32 = 0; int32_t sum_log_magn = 0; int32_t sum_log_i_log_magn = 0; uint16_t sum_log_magn_u16 = 0; uint16_t tmp_u16 = 0; int16_t sum_log_i = 0; int16_t sum_log_i_square = 0; int16_t frac = 0; int16_t log2 = 0; int16_t matrix_determinant = 0; int16_t maxWinData; size_t i, j; int zeros; int net_norm = 0; int right_shifts_in_magnU16 = 0; int right_shifts_in_initMagnEst = 0; int16_t winData_buff[ANAL_BLOCKL_MAX * 2 + 16]; int16_t realImag_buff[ANAL_BLOCKL_MAX * 2 + 16]; // Align the structures to 32-byte boundary for the FFT function. int16_t* winData = (int16_t*) (((uintptr_t)winData_buff + 31) & ~31); int16_t* realImag = (int16_t*) (((uintptr_t) realImag_buff + 31) & ~31); // Update analysis buffer for lower band, and window data before FFT. WebRtcNsx_AnalysisUpdate(inst, winData, speechFrame); // Get input energy inst->energyIn = WebRtcSpl_Energy(winData, inst->anaLen, &inst->scaleEnergyIn); // Reset zero input flag inst->zeroInputSignal = 0; // Acquire norm for winData maxWinData = WebRtcSpl_MaxAbsValueW16(winData, inst->anaLen); inst->normData = WebRtcSpl_NormW16(maxWinData); if (maxWinData == 0) { // Treat zero input separately. inst->zeroInputSignal = 1; return; } // Determine the net normalization in the frequency domain net_norm = inst->stages - inst->normData; // Track lowest normalization factor and use it to prevent wrap around in shifting right_shifts_in_magnU16 = inst->normData - inst->minNorm; right_shifts_in_initMagnEst = WEBRTC_SPL_MAX(-right_shifts_in_magnU16, 0); inst->minNorm -= right_shifts_in_initMagnEst; right_shifts_in_magnU16 = WEBRTC_SPL_MAX(right_shifts_in_magnU16, 0); // create realImag as winData interleaved with zeros (= imag. part), normalize it WebRtcNsx_NormalizeRealBuffer(inst, winData, realImag); // FFT output will be in winData[]. WebRtcSpl_RealForwardFFT(inst->real_fft, realImag, winData); inst->imag[0] = 0; // Q(normData-stages) inst->imag[inst->anaLen2] = 0; inst->real[0] = winData[0]; // Q(normData-stages) inst->real[inst->anaLen2] = winData[inst->anaLen]; // Q(2*(normData-stages)) inst->magnEnergy = (uint32_t)(inst->real[0] * inst->real[0]); inst->magnEnergy += (uint32_t)(inst->real[inst->anaLen2] * inst->real[inst->anaLen2]); magnU16[0] = (uint16_t)WEBRTC_SPL_ABS_W16(inst->real[0]); // Q(normData-stages) magnU16[inst->anaLen2] = (uint16_t)WEBRTC_SPL_ABS_W16(inst->real[inst->anaLen2]); inst->sumMagn = (uint32_t)magnU16[0]; // Q(normData-stages) inst->sumMagn += (uint32_t)magnU16[inst->anaLen2]; if (inst->blockIndex >= END_STARTUP_SHORT) { for (i = 1, j = 2; i < inst->anaLen2; i += 1, j += 2) { inst->real[i] = winData[j]; inst->imag[i] = -winData[j + 1]; // magnitude spectrum // energy in Q(2*(normData-stages)) tmpU32no1 = (uint32_t)(winData[j] * winData[j]); tmpU32no1 += (uint32_t)(winData[j + 1] * winData[j + 1]); inst->magnEnergy += tmpU32no1; // Q(2*(normData-stages)) magnU16[i] = (uint16_t)WebRtcSpl_SqrtFloor(tmpU32no1); // Q(normData-stages) inst->sumMagn += (uint32_t)magnU16[i]; // Q(normData-stages) } } else { // // Gather information during startup for noise parameter estimation // // Switch initMagnEst to Q(minNorm-stages) inst->initMagnEst[0] >>= right_shifts_in_initMagnEst; inst->initMagnEst[inst->anaLen2] >>= right_shifts_in_initMagnEst; // Update initMagnEst with magnU16 in Q(minNorm-stages). inst->initMagnEst[0] += magnU16[0] >> right_shifts_in_magnU16; inst->initMagnEst[inst->anaLen2] += magnU16[inst->anaLen2] >> right_shifts_in_magnU16; log2 = 0; if (magnU16[inst->anaLen2]) { // Calculate log2(magnU16[inst->anaLen2]) zeros = WebRtcSpl_NormU32((uint32_t)magnU16[inst->anaLen2]); frac = (int16_t)((((uint32_t)magnU16[inst->anaLen2] << zeros) & 0x7FFFFFFF) >> 23); // Q8 // log2(magnU16(i)) in Q8 assert(frac < 256); log2 = (int16_t)(((31 - zeros) << 8) + WebRtcNsx_kLogTableFrac[frac]); } sum_log_magn = (int32_t)log2; // Q8 // sum_log_i_log_magn in Q17 sum_log_i_log_magn = (kLogIndex[inst->anaLen2] * log2) >> 3; for (i = 1, j = 2; i < inst->anaLen2; i += 1, j += 2) { inst->real[i] = winData[j]; inst->imag[i] = -winData[j + 1]; // magnitude spectrum // energy in Q(2*(normData-stages)) tmpU32no1 = (uint32_t)(winData[j] * winData[j]); tmpU32no1 += (uint32_t)(winData[j + 1] * winData[j + 1]); inst->magnEnergy += tmpU32no1; // Q(2*(normData-stages)) magnU16[i] = (uint16_t)WebRtcSpl_SqrtFloor(tmpU32no1); // Q(normData-stages) inst->sumMagn += (uint32_t)magnU16[i]; // Q(normData-stages) // Switch initMagnEst to Q(minNorm-stages) inst->initMagnEst[i] >>= right_shifts_in_initMagnEst; // Update initMagnEst with magnU16 in Q(minNorm-stages). inst->initMagnEst[i] += magnU16[i] >> right_shifts_in_magnU16; if (i >= kStartBand) { // For pink noise estimation. Collect data neglecting lower frequency band log2 = 0; if (magnU16[i]) { zeros = WebRtcSpl_NormU32((uint32_t)magnU16[i]); frac = (int16_t)((((uint32_t)magnU16[i] << zeros) & 0x7FFFFFFF) >> 23); // log2(magnU16(i)) in Q8 assert(frac < 256); log2 = (int16_t)(((31 - zeros) << 8) + WebRtcNsx_kLogTableFrac[frac]); } sum_log_magn += (int32_t)log2; // Q8 // sum_log_i_log_magn in Q17 sum_log_i_log_magn += (kLogIndex[i] * log2) >> 3; } } // //compute simplified noise model during startup // // Estimate White noise // Switch whiteNoiseLevel to Q(minNorm-stages) inst->whiteNoiseLevel >>= right_shifts_in_initMagnEst; // Update the average magnitude spectrum, used as noise estimate. tmpU32no1 = WEBRTC_SPL_UMUL_32_16(inst->sumMagn, inst->overdrive); tmpU32no1 >>= inst->stages + 8; // Replacing division above with 'stages' shifts // Shift to same Q-domain as whiteNoiseLevel tmpU32no1 >>= right_shifts_in_magnU16; // This operation is safe from wrap around as long as END_STARTUP_SHORT < 128 assert(END_STARTUP_SHORT < 128); inst->whiteNoiseLevel += tmpU32no1; // Q(minNorm-stages) // Estimate Pink noise parameters // Denominator used in both parameter estimates. // The value is only dependent on the size of the frequency band (kStartBand) // and to reduce computational complexity stored in a table (kDeterminantEstMatrix[]) assert(kStartBand < 66); matrix_determinant = kDeterminantEstMatrix[kStartBand]; // Q0 sum_log_i = kSumLogIndex[kStartBand]; // Q5 sum_log_i_square = kSumSquareLogIndex[kStartBand]; // Q2 if (inst->fs == 8000) { // Adjust values to shorter blocks in narrow band. tmp_1_w32 = (int32_t)matrix_determinant; tmp_1_w32 += (kSumLogIndex[65] * sum_log_i) >> 9; tmp_1_w32 -= (kSumLogIndex[65] * kSumLogIndex[65]) >> 10; tmp_1_w32 -= (int32_t)sum_log_i_square << 4; tmp_1_w32 -= ((inst->magnLen - kStartBand) * kSumSquareLogIndex[65]) >> 2; matrix_determinant = (int16_t)tmp_1_w32; sum_log_i -= kSumLogIndex[65]; // Q5 sum_log_i_square -= kSumSquareLogIndex[65]; // Q2 } // Necessary number of shifts to fit sum_log_magn in a word16 zeros = 16 - WebRtcSpl_NormW32(sum_log_magn); if (zeros < 0) { zeros = 0; } tmp_1_w32 = sum_log_magn << 1; // Q9 sum_log_magn_u16 = (uint16_t)(tmp_1_w32 >> zeros); // Q(9-zeros). // Calculate and update pinkNoiseNumerator. Result in Q11. tmp_2_w32 = WEBRTC_SPL_MUL_16_U16(sum_log_i_square, sum_log_magn_u16); // Q(11-zeros) tmpU32no1 = sum_log_i_log_magn >> 12; // Q5 // Shift the largest value of sum_log_i and tmp32no3 before multiplication tmp_u16 = ((uint16_t)sum_log_i << 1); // Q6 if ((uint32_t)sum_log_i > tmpU32no1) { tmp_u16 >>= zeros; } else { tmpU32no1 >>= zeros; } tmp_2_w32 -= (int32_t)WEBRTC_SPL_UMUL_32_16(tmpU32no1, tmp_u16); // Q(11-zeros) matrix_determinant >>= zeros; // Q(-zeros) tmp_2_w32 = WebRtcSpl_DivW32W16(tmp_2_w32, matrix_determinant); // Q11 tmp_2_w32 += (int32_t)net_norm << 11; // Q11 if (tmp_2_w32 < 0) { tmp_2_w32 = 0; } inst->pinkNoiseNumerator += tmp_2_w32; // Q11 // Calculate and update pinkNoiseExp. Result in Q14. tmp_2_w32 = WEBRTC_SPL_MUL_16_U16(sum_log_i, sum_log_magn_u16); // Q(14-zeros) tmp_1_w32 = sum_log_i_log_magn >> (3 + zeros); tmp_1_w32 *= inst->magnLen - kStartBand; tmp_2_w32 -= tmp_1_w32; // Q(14-zeros) if (tmp_2_w32 > 0) { // If the exponential parameter is negative force it to zero, which means a // flat spectrum. tmp_1_w32 = WebRtcSpl_DivW32W16(tmp_2_w32, matrix_determinant); // Q14 inst->pinkNoiseExp += WEBRTC_SPL_SAT(16384, tmp_1_w32, 0); // Q14 } } } void WebRtcNsx_DataSynthesis(NoiseSuppressionFixedC* inst, short* outFrame) { int32_t energyOut; int16_t realImag_buff[ANAL_BLOCKL_MAX * 2 + 16]; int16_t rfft_out_buff[ANAL_BLOCKL_MAX * 2 + 16]; // Align the structures to 32-byte boundary for the FFT function. int16_t* realImag = (int16_t*) (((uintptr_t)realImag_buff + 31) & ~31); int16_t* rfft_out = (int16_t*) (((uintptr_t) rfft_out_buff + 31) & ~31); int16_t tmp16no1, tmp16no2; int16_t energyRatio; int16_t gainFactor, gainFactor1, gainFactor2; size_t i; int outCIFFT; int scaleEnergyOut = 0; if (inst->zeroInputSignal) { // synthesize the special case of zero input // read out fully processed segment for (i = 0; i < inst->blockLen10ms; i++) { outFrame[i] = inst->synthesisBuffer[i]; // Q0 } // update synthesis buffer memcpy(inst->synthesisBuffer, inst->synthesisBuffer + inst->blockLen10ms, (inst->anaLen - inst->blockLen10ms) * sizeof(*inst->synthesisBuffer)); WebRtcSpl_ZerosArrayW16(inst->synthesisBuffer + inst->anaLen - inst->blockLen10ms, inst->blockLen10ms); return; } // Filter the data in the frequency domain, and create spectrum. WebRtcNsx_PrepareSpectrum(inst, realImag); // Inverse FFT output will be in rfft_out[]. outCIFFT = WebRtcSpl_RealInverseFFT(inst->real_fft, realImag, rfft_out); WebRtcNsx_Denormalize(inst, rfft_out, outCIFFT); //scale factor: only do it after END_STARTUP_LONG time gainFactor = 8192; // 8192 = Q13(1.0) if (inst->gainMap == 1 && inst->blockIndex > END_STARTUP_LONG && inst->energyIn > 0) { // Q(-scaleEnergyOut) energyOut = WebRtcSpl_Energy(inst->real, inst->anaLen, &scaleEnergyOut); if (scaleEnergyOut == 0 && !(energyOut & 0x7f800000)) { energyOut = WEBRTC_SPL_SHIFT_W32(energyOut, 8 + scaleEnergyOut - inst->scaleEnergyIn); } else { // |energyIn| is currently in Q(|scaleEnergyIn|), but to later on end up // with an |energyRatio| in Q8 we need to change the Q-domain to // Q(-8-scaleEnergyOut). inst->energyIn >>= 8 + scaleEnergyOut - inst->scaleEnergyIn; } assert(inst->energyIn > 0); energyRatio = (energyOut + inst->energyIn / 2) / inst->energyIn; // Q8 // Limit the ratio to [0, 1] in Q8, i.e., [0, 256] energyRatio = WEBRTC_SPL_SAT(256, energyRatio, 0); // all done in lookup tables now assert(energyRatio < 257); gainFactor1 = kFactor1Table[energyRatio]; // Q8 gainFactor2 = inst->factor2Table[energyRatio]; // Q8 //combine both scales with speech/noise prob: note prior (priorSpeechProb) is not frequency dependent // factor = inst->priorSpeechProb*factor1 + (1.0-inst->priorSpeechProb)*factor2; // original code tmp16no1 = (int16_t)(((16384 - inst->priorNonSpeechProb) * gainFactor1) >> 14); // in Q13, where 16384 = Q14(1.0) tmp16no2 = (int16_t)((inst->priorNonSpeechProb * gainFactor2) >> 14); gainFactor = tmp16no1 + tmp16no2; // Q13 } // out of flag_gain_map==1 // Synthesis, read out fully processed segment, and update synthesis buffer. WebRtcNsx_SynthesisUpdate(inst, outFrame, gainFactor); } void WebRtcNsx_ProcessCore(NoiseSuppressionFixedC* inst, const short* const* speechFrame, int num_bands, short* const* outFrame) { // main routine for noise suppression uint32_t tmpU32no1, tmpU32no2, tmpU32no3; uint32_t satMax, maxNoiseU32; uint32_t tmpMagnU32, tmpNoiseU32; uint32_t nearMagnEst; uint32_t noiseUpdateU32; uint32_t noiseU32[HALF_ANAL_BLOCKL]; uint32_t postLocSnr[HALF_ANAL_BLOCKL]; uint32_t priorLocSnr[HALF_ANAL_BLOCKL]; uint32_t prevNearSnr[HALF_ANAL_BLOCKL]; uint32_t curNearSnr; uint32_t priorSnr; uint32_t noise_estimate = 0; uint32_t noise_estimate_avg = 0; uint32_t numerator = 0; int32_t tmp32no1, tmp32no2; int32_t pink_noise_num_avg = 0; uint16_t tmpU16no1; uint16_t magnU16[HALF_ANAL_BLOCKL]; uint16_t prevNoiseU16[HALF_ANAL_BLOCKL]; uint16_t nonSpeechProbFinal[HALF_ANAL_BLOCKL]; uint16_t gammaNoise, prevGammaNoise; uint16_t noiseSupFilterTmp[HALF_ANAL_BLOCKL]; int16_t qMagn, qNoise; int16_t avgProbSpeechHB, gainModHB, avgFilterGainHB, gainTimeDomainHB; int16_t pink_noise_exp_avg = 0; size_t i, j; int nShifts, postShifts; int norm32no1, norm32no2; int flag, sign; int q_domain_to_use = 0; // Code for ARMv7-Neon platform assumes the following: assert(inst->anaLen > 0); assert(inst->anaLen2 > 0); assert(inst->anaLen % 16 == 0); assert(inst->anaLen2 % 8 == 0); assert(inst->blockLen10ms > 0); assert(inst->blockLen10ms % 16 == 0); assert(inst->magnLen == inst->anaLen2 + 1); #ifdef NS_FILEDEBUG if (fwrite(spframe, sizeof(short), inst->blockLen10ms, inst->infile) != inst->blockLen10ms) { assert(false); } #endif // Check that initialization has been done assert(inst->initFlag == 1); assert((num_bands - 1) <= NUM_HIGH_BANDS_MAX); const short* const* speechFrameHB = NULL; short* const* outFrameHB = NULL; size_t num_high_bands = 0; if (num_bands > 1) { speechFrameHB = &speechFrame[1]; outFrameHB = &outFrame[1]; num_high_bands = (size_t)(num_bands - 1); } // Store speechFrame and transform to frequency domain WebRtcNsx_DataAnalysis(inst, (short*)speechFrame[0], magnU16); if (inst->zeroInputSignal) { WebRtcNsx_DataSynthesis(inst, outFrame[0]); if (num_bands > 1) { // update analysis buffer for H band // append new data to buffer FX for (i = 0; i < num_high_bands; ++i) { int block_shift = inst->anaLen - inst->blockLen10ms; memcpy(inst->dataBufHBFX[i], inst->dataBufHBFX[i] + inst->blockLen10ms, block_shift * sizeof(*inst->dataBufHBFX[i])); memcpy(inst->dataBufHBFX[i] + block_shift, speechFrameHB[i], inst->blockLen10ms * sizeof(*inst->dataBufHBFX[i])); for (j = 0; j < inst->blockLen10ms; j++) { outFrameHB[i][j] = inst->dataBufHBFX[i][j]; // Q0 } } } // end of H band gain computation return; } // Update block index when we have something to process inst->blockIndex++; // // Norm of magn qMagn = inst->normData - inst->stages; // Compute spectral flatness on input spectrum WebRtcNsx_ComputeSpectralFlatness(inst, magnU16); // quantile noise estimate WebRtcNsx_NoiseEstimation(inst, magnU16, noiseU32, &qNoise); //noise estimate from previous frame for (i = 0; i < inst->magnLen; i++) { prevNoiseU16[i] = (uint16_t)(inst->prevNoiseU32[i] >> 11); // Q(prevQNoise) } if (inst->blockIndex < END_STARTUP_SHORT) { // Noise Q-domain to be used later; see description at end of section. q_domain_to_use = WEBRTC_SPL_MIN((int)qNoise, inst->minNorm - inst->stages); // Calculate frequency independent parts in parametric noise estimate and calculate // the estimate for the lower frequency band (same values for all frequency bins) if (inst->pinkNoiseExp) { pink_noise_exp_avg = (int16_t)WebRtcSpl_DivW32W16(inst->pinkNoiseExp, (int16_t)(inst->blockIndex + 1)); // Q14 pink_noise_num_avg = WebRtcSpl_DivW32W16(inst->pinkNoiseNumerator, (int16_t)(inst->blockIndex + 1)); // Q11 WebRtcNsx_CalcParametricNoiseEstimate(inst, pink_noise_exp_avg, pink_noise_num_avg, kStartBand, &noise_estimate, &noise_estimate_avg); } else { // Use white noise estimate if we have poor pink noise parameter estimates noise_estimate = inst->whiteNoiseLevel; // Q(minNorm-stages) noise_estimate_avg = noise_estimate / (inst->blockIndex + 1); // Q(minNorm-stages) } for (i = 0; i < inst->magnLen; i++) { // Estimate the background noise using the pink noise parameters if permitted if ((inst->pinkNoiseExp) && (i >= kStartBand)) { // Reset noise_estimate noise_estimate = 0; noise_estimate_avg = 0; // Calculate the parametric noise estimate for current frequency bin WebRtcNsx_CalcParametricNoiseEstimate(inst, pink_noise_exp_avg, pink_noise_num_avg, i, &noise_estimate, &noise_estimate_avg); } // Calculate parametric Wiener filter noiseSupFilterTmp[i] = inst->denoiseBound; if (inst->initMagnEst[i]) { // numerator = (initMagnEst - noise_estimate * overdrive) // Result in Q(8+minNorm-stages) tmpU32no1 = WEBRTC_SPL_UMUL_32_16(noise_estimate, inst->overdrive); numerator = inst->initMagnEst[i] << 8; if (numerator > tmpU32no1) { // Suppression filter coefficient larger than zero, so calculate. numerator -= tmpU32no1; // Determine number of left shifts in numerator for best accuracy after // division nShifts = WebRtcSpl_NormU32(numerator); nShifts = WEBRTC_SPL_SAT(6, nShifts, 0); // Shift numerator to Q(nShifts+8+minNorm-stages) numerator <<= nShifts; // Shift denominator to Q(nShifts-6+minNorm-stages) tmpU32no1 = inst->initMagnEst[i] >> (6 - nShifts); if (tmpU32no1 == 0) { // This is only possible if numerator = 0, in which case // we don't need any division. tmpU32no1 = 1; } tmpU32no2 = numerator / tmpU32no1; // Q14 noiseSupFilterTmp[i] = (uint16_t)WEBRTC_SPL_SAT(16384, tmpU32no2, (uint32_t)(inst->denoiseBound)); // Q14 } } // Weight quantile noise 'noiseU32' with modeled noise 'noise_estimate_avg' // 'noiseU32 is in Q(qNoise) and 'noise_estimate' in Q(minNorm-stages) // To guarantee that we do not get wrap around when shifting to the same domain // we use the lowest one. Furthermore, we need to save 6 bits for the weighting. // 'noise_estimate_avg' can handle this operation by construction, but 'noiseU32' // may not. // Shift 'noiseU32' to 'q_domain_to_use' tmpU32no1 = noiseU32[i] >> (qNoise - q_domain_to_use); // Shift 'noise_estimate_avg' to 'q_domain_to_use' tmpU32no2 = noise_estimate_avg >> (inst->minNorm - inst->stages - q_domain_to_use); // Make a simple check to see if we have enough room for weighting 'tmpU32no1' // without wrap around nShifts = 0; if (tmpU32no1 & 0xfc000000) { tmpU32no1 >>= 6; tmpU32no2 >>= 6; nShifts = 6; } tmpU32no1 *= inst->blockIndex; tmpU32no2 *= (END_STARTUP_SHORT - inst->blockIndex); // Add them together and divide by startup length noiseU32[i] = WebRtcSpl_DivU32U16(tmpU32no1 + tmpU32no2, END_STARTUP_SHORT); // Shift back if necessary noiseU32[i] <<= nShifts; } // Update new Q-domain for 'noiseU32' qNoise = q_domain_to_use; } // compute average signal during END_STARTUP_LONG time: // used to normalize spectral difference measure if (inst->blockIndex < END_STARTUP_LONG) { // substituting division with shift ending up in Q(-2*stages) inst->timeAvgMagnEnergyTmp += inst->magnEnergy >> (2 * inst->normData + inst->stages - 1); inst->timeAvgMagnEnergy = WebRtcSpl_DivU32U16(inst->timeAvgMagnEnergyTmp, inst->blockIndex + 1); } //start processing at frames == converged+1 // STEP 1: compute prior and post SNR based on quantile noise estimates // compute direct decision (DD) estimate of prior SNR: needed for new method satMax = (uint32_t)1048575;// Largest possible value without getting overflow despite shifting 12 steps postShifts = 6 + qMagn - qNoise; nShifts = 5 - inst->prevQMagn + inst->prevQNoise; for (i = 0; i < inst->magnLen; i++) { // FLOAT: // post SNR // postLocSnr[i] = 0.0; // if (magn[i] > noise[i]) // { // postLocSnr[i] = magn[i] / (noise[i] + 0.0001); // } // // previous post SNR // // previous estimate: based on previous frame with gain filter (smooth is previous filter) // // prevNearSnr[i] = inst->prevMagnU16[i] / (inst->noisePrev[i] + 0.0001) * (inst->smooth[i]); // // // DD estimate is sum of two terms: current estimate and previous estimate // // directed decision update of priorSnr (or we actually store [2*priorSnr+1]) // // priorLocSnr[i] = DD_PR_SNR * prevNearSnr[i] + (1.0 - DD_PR_SNR) * (postLocSnr[i] - 1.0); // calculate post SNR: output in Q11 postLocSnr[i] = 2048; // 1.0 in Q11 tmpU32no1 = (uint32_t)magnU16[i] << 6; // Q(6+qMagn) if (postShifts < 0) { tmpU32no2 = noiseU32[i] >> -postShifts; // Q(6+qMagn) } else { tmpU32no2 = noiseU32[i] << postShifts; // Q(6+qMagn) } if (tmpU32no1 > tmpU32no2) { // Current magnitude larger than noise tmpU32no1 <<= 11; // Q(17+qMagn) if (tmpU32no2 > 0) { tmpU32no1 /= tmpU32no2; // Q11 postLocSnr[i] = WEBRTC_SPL_MIN(satMax, tmpU32no1); // Q11 } else { postLocSnr[i] = satMax; } } // calculate prevNearSnr[i] and save for later instead of recalculating it later // |nearMagnEst| in Q(prevQMagn + 14) nearMagnEst = inst->prevMagnU16[i] * inst->noiseSupFilter[i]; tmpU32no1 = nearMagnEst << 3; // Q(prevQMagn+17) tmpU32no2 = inst->prevNoiseU32[i] >> nShifts; // Q(prevQMagn+6) if (tmpU32no2 > 0) { tmpU32no1 /= tmpU32no2; // Q11 tmpU32no1 = WEBRTC_SPL_MIN(satMax, tmpU32no1); // Q11 } else { tmpU32no1 = satMax; // Q11 } prevNearSnr[i] = tmpU32no1; // Q11 //directed decision update of priorSnr tmpU32no1 = WEBRTC_SPL_UMUL_32_16(prevNearSnr[i], DD_PR_SNR_Q11); // Q22 tmpU32no2 = WEBRTC_SPL_UMUL_32_16(postLocSnr[i] - 2048, ONE_MINUS_DD_PR_SNR_Q11); // Q22 priorSnr = tmpU32no1 + tmpU32no2 + 512; // Q22 (added 512 for rounding) // priorLocSnr = 1 + 2*priorSnr priorLocSnr[i] = 2048 + (priorSnr >> 10); // Q11 } // end of loop over frequencies // done with step 1: DD computation of prior and post SNR // STEP 2: compute speech/noise likelihood //compute difference of input spectrum with learned/estimated noise spectrum WebRtcNsx_ComputeSpectralDifference(inst, magnU16); //compute histograms for determination of parameters (thresholds and weights for features) //parameters are extracted once every window time (=inst->modelUpdate) //counter update inst->cntThresUpdate++; flag = (int)(inst->cntThresUpdate == inst->modelUpdate); //update histogram WebRtcNsx_FeatureParameterExtraction(inst, flag); //compute model parameters if (flag) { inst->cntThresUpdate = 0; // Reset counter //update every window: // get normalization for spectral difference for next window estimate // Shift to Q(-2*stages) inst->curAvgMagnEnergy >>= STAT_UPDATES; tmpU32no1 = (inst->curAvgMagnEnergy + inst->timeAvgMagnEnergy + 1) >> 1; //Q(-2*stages) // Update featureSpecDiff if ((tmpU32no1 != inst->timeAvgMagnEnergy) && (inst->featureSpecDiff) && (inst->timeAvgMagnEnergy > 0)) { norm32no1 = 0; tmpU32no3 = tmpU32no1; while (0xFFFF0000 & tmpU32no3) { tmpU32no3 >>= 1; norm32no1++; } tmpU32no2 = inst->featureSpecDiff; while (0xFFFF0000 & tmpU32no2) { tmpU32no2 >>= 1; norm32no1++; } tmpU32no3 = WEBRTC_SPL_UMUL(tmpU32no3, tmpU32no2); tmpU32no3 /= inst->timeAvgMagnEnergy; if (WebRtcSpl_NormU32(tmpU32no3) < norm32no1) { inst->featureSpecDiff = 0x007FFFFF; } else { inst->featureSpecDiff = WEBRTC_SPL_MIN(0x007FFFFF, tmpU32no3 << norm32no1); } } inst->timeAvgMagnEnergy = tmpU32no1; // Q(-2*stages) inst->curAvgMagnEnergy = 0; } //compute speech/noise probability WebRtcNsx_SpeechNoiseProb(inst, nonSpeechProbFinal, priorLocSnr, postLocSnr); //time-avg parameter for noise update gammaNoise = NOISE_UPDATE_Q8; // Q8 maxNoiseU32 = 0; postShifts = inst->prevQNoise - qMagn; nShifts = inst->prevQMagn - qMagn; for (i = 0; i < inst->magnLen; i++) { // temporary noise update: use it for speech frames if update value is less than previous // the formula has been rewritten into: // noiseUpdate = noisePrev[i] + (1 - gammaNoise) * nonSpeechProb * (magn[i] - noisePrev[i]) if (postShifts < 0) { tmpU32no2 = magnU16[i] >> -postShifts; // Q(prevQNoise) } else { tmpU32no2 = (uint32_t)magnU16[i] << postShifts; // Q(prevQNoise) } if (prevNoiseU16[i] > tmpU32no2) { sign = -1; tmpU32no1 = prevNoiseU16[i] - tmpU32no2; } else { sign = 1; tmpU32no1 = tmpU32no2 - prevNoiseU16[i]; } noiseUpdateU32 = inst->prevNoiseU32[i]; // Q(prevQNoise+11) tmpU32no3 = 0; if ((tmpU32no1) && (nonSpeechProbFinal[i])) { // This value will be used later, if gammaNoise changes tmpU32no3 = WEBRTC_SPL_UMUL_32_16(tmpU32no1, nonSpeechProbFinal[i]); // Q(prevQNoise+8) if (0x7c000000 & tmpU32no3) { // Shifting required before multiplication tmpU32no2 = (tmpU32no3 >> 5) * gammaNoise; // Q(prevQNoise+11) } else { // We can do shifting after multiplication tmpU32no2 = (tmpU32no3 * gammaNoise) >> 5; // Q(prevQNoise+11) } if (sign > 0) { noiseUpdateU32 += tmpU32no2; // Q(prevQNoise+11) } else { // This operation is safe. We can never get wrap around, since worst // case scenario means magnU16 = 0 noiseUpdateU32 -= tmpU32no2; // Q(prevQNoise+11) } } //increase gamma (i.e., less noise update) for frame likely to be speech prevGammaNoise = gammaNoise; gammaNoise = NOISE_UPDATE_Q8; //time-constant based on speech/noise state //increase gamma (i.e., less noise update) for frames likely to be speech if (nonSpeechProbFinal[i] < ONE_MINUS_PROB_RANGE_Q8) { gammaNoise = GAMMA_NOISE_TRANS_AND_SPEECH_Q8; } if (prevGammaNoise != gammaNoise) { // new noise update // this line is the same as above, only that the result is stored in a different variable and the gammaNoise // has changed // // noiseUpdate = noisePrev[i] + (1 - gammaNoise) * nonSpeechProb * (magn[i] - noisePrev[i]) if (0x7c000000 & tmpU32no3) { // Shifting required before multiplication tmpU32no2 = (tmpU32no3 >> 5) * gammaNoise; // Q(prevQNoise+11) } else { // We can do shifting after multiplication tmpU32no2 = (tmpU32no3 * gammaNoise) >> 5; // Q(prevQNoise+11) } if (sign > 0) { tmpU32no1 = inst->prevNoiseU32[i] + tmpU32no2; // Q(prevQNoise+11) } else { tmpU32no1 = inst->prevNoiseU32[i] - tmpU32no2; // Q(prevQNoise+11) } if (noiseUpdateU32 > tmpU32no1) { noiseUpdateU32 = tmpU32no1; // Q(prevQNoise+11) } } noiseU32[i] = noiseUpdateU32; // Q(prevQNoise+11) if (noiseUpdateU32 > maxNoiseU32) { maxNoiseU32 = noiseUpdateU32; } // conservative noise update // // original FLOAT code // if (prob_speech < PROB_RANGE) { // inst->avgMagnPause[i] = inst->avgMagnPause[i] + (1.0 - gamma_pause)*(magn[i] - inst->avgMagnPause[i]); // } tmp32no2 = WEBRTC_SPL_SHIFT_W32(inst->avgMagnPause[i], -nShifts); if (nonSpeechProbFinal[i] > ONE_MINUS_PROB_RANGE_Q8) { if (nShifts < 0) { tmp32no1 = (int32_t)magnU16[i] - tmp32no2; // Q(qMagn) tmp32no1 *= ONE_MINUS_GAMMA_PAUSE_Q8; // Q(8+prevQMagn+nShifts) tmp32no1 = (tmp32no1 + 128) >> 8; // Q(qMagn). } else { // In Q(qMagn+nShifts) tmp32no1 = ((int32_t)magnU16[i] << nShifts) - inst->avgMagnPause[i]; tmp32no1 *= ONE_MINUS_GAMMA_PAUSE_Q8; // Q(8+prevQMagn+nShifts) tmp32no1 = (tmp32no1 + (128 << nShifts)) >> (8 + nShifts); // Q(qMagn). } tmp32no2 += tmp32no1; // Q(qMagn) } inst->avgMagnPause[i] = tmp32no2; } // end of frequency loop norm32no1 = WebRtcSpl_NormU32(maxNoiseU32); qNoise = inst->prevQNoise + norm32no1 - 5; // done with step 2: noise update // STEP 3: compute dd update of prior snr and post snr based on new noise estimate nShifts = inst->prevQNoise + 11 - qMagn; for (i = 0; i < inst->magnLen; i++) { // FLOAT code // // post and prior SNR // curNearSnr = 0.0; // if (magn[i] > noise[i]) // { // curNearSnr = magn[i] / (noise[i] + 0.0001) - 1.0; // } // // DD estimate is sum of two terms: current estimate and previous estimate // // directed decision update of snrPrior // snrPrior = DD_PR_SNR * prevNearSnr[i] + (1.0 - DD_PR_SNR) * curNearSnr; // // gain filter // tmpFloat1 = inst->overdrive + snrPrior; // tmpFloat2 = snrPrior / tmpFloat1; // theFilter[i] = tmpFloat2; // calculate curNearSnr again, this is necessary because a new noise estimate has been made since then. for the original curNearSnr = 0; // Q11 if (nShifts < 0) { // This case is equivalent with magn < noise which implies curNearSnr = 0; tmpMagnU32 = (uint32_t)magnU16[i]; // Q(qMagn) tmpNoiseU32 = noiseU32[i] << -nShifts; // Q(qMagn) } else if (nShifts > 17) { tmpMagnU32 = (uint32_t)magnU16[i] << 17; // Q(qMagn+17) tmpNoiseU32 = noiseU32[i] >> (nShifts - 17); // Q(qMagn+17) } else { tmpMagnU32 = (uint32_t)magnU16[i] << nShifts; // Q(qNoise_prev+11) tmpNoiseU32 = noiseU32[i]; // Q(qNoise_prev+11) } if (tmpMagnU32 > tmpNoiseU32) { tmpU32no1 = tmpMagnU32 - tmpNoiseU32; // Q(qCur) norm32no2 = WEBRTC_SPL_MIN(11, WebRtcSpl_NormU32(tmpU32no1)); tmpU32no1 <<= norm32no2; // Q(qCur+norm32no2) tmpU32no2 = tmpNoiseU32 >> (11 - norm32no2); // Q(qCur+norm32no2-11) if (tmpU32no2 > 0) { tmpU32no1 /= tmpU32no2; // Q11 } curNearSnr = WEBRTC_SPL_MIN(satMax, tmpU32no1); // Q11 } //directed decision update of priorSnr // FLOAT // priorSnr = DD_PR_SNR * prevNearSnr + (1.0-DD_PR_SNR) * curNearSnr; tmpU32no1 = WEBRTC_SPL_UMUL_32_16(prevNearSnr[i], DD_PR_SNR_Q11); // Q22 tmpU32no2 = WEBRTC_SPL_UMUL_32_16(curNearSnr, ONE_MINUS_DD_PR_SNR_Q11); // Q22 priorSnr = tmpU32no1 + tmpU32no2; // Q22 //gain filter tmpU32no1 = inst->overdrive + ((priorSnr + 8192) >> 14); // Q8 assert(inst->overdrive > 0); tmpU16no1 = (priorSnr + tmpU32no1 / 2) / tmpU32no1; // Q14 inst->noiseSupFilter[i] = WEBRTC_SPL_SAT(16384, tmpU16no1, inst->denoiseBound); // 16384 = Q14(1.0) // Q14 // Weight in the parametric Wiener filter during startup if (inst->blockIndex < END_STARTUP_SHORT) { // Weight the two suppression filters tmpU32no1 = inst->noiseSupFilter[i] * inst->blockIndex; tmpU32no2 = noiseSupFilterTmp[i] * (END_STARTUP_SHORT - inst->blockIndex); tmpU32no1 += tmpU32no2; inst->noiseSupFilter[i] = (uint16_t)WebRtcSpl_DivU32U16(tmpU32no1, END_STARTUP_SHORT); } } // end of loop over frequencies //done with step3 // save noise and magnitude spectrum for next frame inst->prevQNoise = qNoise; inst->prevQMagn = qMagn; if (norm32no1 > 5) { for (i = 0; i < inst->magnLen; i++) { inst->prevNoiseU32[i] = noiseU32[i] << (norm32no1 - 5); // Q(qNoise+11) inst->prevMagnU16[i] = magnU16[i]; // Q(qMagn) } } else { for (i = 0; i < inst->magnLen; i++) { inst->prevNoiseU32[i] = noiseU32[i] >> (5 - norm32no1); // Q(qNoise+11) inst->prevMagnU16[i] = magnU16[i]; // Q(qMagn) } } WebRtcNsx_DataSynthesis(inst, outFrame[0]); #ifdef NS_FILEDEBUG if (fwrite(outframe, sizeof(short), inst->blockLen10ms, inst->outfile) != inst->blockLen10ms) { assert(false); } #endif //for H band: // only update data buffer, then apply time-domain gain is applied derived from L band if (num_bands > 1) { // update analysis buffer for H band // append new data to buffer FX for (i = 0; i < num_high_bands; ++i) { memcpy(inst->dataBufHBFX[i], inst->dataBufHBFX[i] + inst->blockLen10ms, (inst->anaLen - inst->blockLen10ms) * sizeof(*inst->dataBufHBFX[i])); memcpy(inst->dataBufHBFX[i] + inst->anaLen - inst->blockLen10ms, speechFrameHB[i], inst->blockLen10ms * sizeof(*inst->dataBufHBFX[i])); } // range for averaging low band quantities for H band gain gainTimeDomainHB = 16384; // 16384 = Q14(1.0) //average speech prob from low band //average filter gain from low band //avg over second half (i.e., 4->8kHz) of freq. spectrum tmpU32no1 = 0; // Q12 tmpU16no1 = 0; // Q8 for (i = inst->anaLen2 - (inst->anaLen2 >> 2); i < inst->anaLen2; i++) { tmpU16no1 += nonSpeechProbFinal[i]; // Q8 tmpU32no1 += (uint32_t)(inst->noiseSupFilter[i]); // Q14 } assert(inst->stages >= 7); avgProbSpeechHB = (4096 - (tmpU16no1 >> (inst->stages - 7))); // Q12 avgFilterGainHB = (int16_t)(tmpU32no1 >> (inst->stages - 3)); // Q14 // // original FLOAT code // // gain based on speech probability: // avg_prob_speech_tt=(float)2.0*avg_prob_speech-(float)1.0; // gain_mod=(float)0.5*((float)1.0+(float)tanh(avg_prob_speech_tt)); // between 0 and 1 // gain based on speech probability: // original expression: "0.5 * (1 + tanh(2x-1))" // avgProbSpeechHB has been anyway saturated to a value between 0 and 1 so the other cases don't have to be dealt with // avgProbSpeechHB and gainModHB are in Q12, 3607 = Q12(0.880615234375) which is a zero point of // |0.5 * (1 + tanh(2x-1)) - x| - |0.5 * (1 + tanh(2x-1)) - 0.880615234375| meaning that from that point the error of approximating // the expression with f(x) = x would be greater than the error of approximating the expression with f(x) = 0.880615234375 // error: "|0.5 * (1 + tanh(2x-1)) - x| from x=0 to 0.880615234375" -> http://www.wolframalpha.com/input/?i=|0.5+*+(1+%2B+tanh(2x-1))+-+x|+from+x%3D0+to+0.880615234375 // and: "|0.5 * (1 + tanh(2x-1)) - 0.880615234375| from x=0.880615234375 to 1" -> http://www.wolframalpha.com/input/?i=+|0.5+*+(1+%2B+tanh(2x-1))+-+0.880615234375|+from+x%3D0.880615234375+to+1 gainModHB = WEBRTC_SPL_MIN(avgProbSpeechHB, 3607); // // original FLOAT code // //combine gain with low band gain // if (avg_prob_speech < (float)0.5) { // gain_time_domain_HB=(float)0.5*gain_mod+(float)0.5*avg_filter_gain; // } // else { // gain_time_domain_HB=(float)0.25*gain_mod+(float)0.75*avg_filter_gain; // } //combine gain with low band gain if (avgProbSpeechHB < 2048) { // 2048 = Q12(0.5) // the next two lines in float are "gain_time_domain = 0.5 * gain_mod + 0.5 * avg_filter_gain"; Q2(0.5) = 2 equals one left shift gainTimeDomainHB = (gainModHB << 1) + (avgFilterGainHB >> 1); // Q14 } else { // "gain_time_domain = 0.25 * gain_mod + 0.75 * agv_filter_gain;" gainTimeDomainHB = (int16_t)((3 * avgFilterGainHB) >> 2); // 3 = Q2(0.75) gainTimeDomainHB += gainModHB; // Q14 } //make sure gain is within flooring range gainTimeDomainHB = WEBRTC_SPL_SAT(16384, gainTimeDomainHB, (int16_t)(inst->denoiseBound)); // 16384 = Q14(1.0) //apply gain for (i = 0; i < num_high_bands; ++i) { for (j = 0; j < inst->blockLen10ms; j++) { outFrameHB[i][j] = (int16_t)((gainTimeDomainHB * inst->dataBufHBFX[i][j]) >> 14); // Q0 } } } // end of H band gain computation }