OLD | NEW |
1 /* | 1 /* |
2 * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. | 2 * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license | 4 * Use of this source code is governed by a BSD-style license |
5 * that can be found in the LICENSE file in the root of the source | 5 * that can be found in the LICENSE file in the root of the source |
6 * tree. An additional intellectual property rights grant can be found | 6 * tree. An additional intellectual property rights grant can be found |
7 * in the file PATENTS. All contributing project authors may | 7 * in the file PATENTS. All contributing project authors may |
8 * be found in the AUTHORS file in the root of the source tree. | 8 * be found in the AUTHORS file in the root of the source tree. |
9 */ | 9 */ |
10 | 10 |
11 // | |
12 // Command line tool for speech intelligibility enhancement. Provides for | |
13 // running and testing intelligibility_enhancer as an independent process. | |
14 // Use --help for options. | |
15 // | |
16 | |
17 #include <arpa/inet.h> | 11 #include <arpa/inet.h> |
18 #include <fcntl.h> | 12 #include <fcntl.h> |
19 #include <stdint.h> | 13 #include <stdint.h> |
20 #include <stdio.h> | 14 #include <stdio.h> |
21 #include <stdlib.h> | 15 #include <stdlib.h> |
22 #include <sys/mman.h> | 16 #include <sys/mman.h> |
23 #include <sys/stat.h> | 17 #include <sys/stat.h> |
24 #include <sys/types.h> | 18 #include <sys/types.h> |
25 #include <unistd.h> | 19 #include <unistd.h> |
26 | 20 |
27 #include <fenv.h> | 21 #include <fenv.h> |
28 #include <limits> | 22 #include <limits> |
29 | 23 |
30 #include <complex> | 24 #include <complex> |
31 | 25 |
32 #include "gflags/gflags.h" | 26 #include "gflags/gflags.h" |
33 #include "testing/gtest/include/gtest/gtest.h" | |
34 #include "webrtc/base/checks.h" | 27 #include "webrtc/base/checks.h" |
35 #include "webrtc/common_audio/real_fourier.h" | 28 #include "webrtc/common_audio/real_fourier.h" |
36 #include "webrtc/modules/audio_processing/intelligibility/intelligibility_enhanc
er.h" | 29 #include "webrtc/modules/audio_processing/intelligibility/intelligibility_enhanc
er.h" |
37 #include "webrtc/modules/audio_processing/intelligibility/intelligibility_utils.
h" | 30 #include "webrtc/modules/audio_processing/intelligibility/intelligibility_utils.
h" |
38 #include "webrtc/system_wrappers/interface/critical_section_wrapper.h" | 31 #include "webrtc/system_wrappers/interface/critical_section_wrapper.h" |
| 32 #include "webrtc/system_wrappers/interface/scoped_ptr.h" |
39 | 33 |
40 // PCM data simulating streams | |
41 const int16_t* in_ipcm; | 34 const int16_t* in_ipcm; |
42 int16_t* out_ipcm; | 35 int16_t* out_ipcm; |
43 const int16_t* noise_ipcm; | 36 const int16_t* noise_ipcm; |
| 37 |
44 float* in_fpcm; | 38 float* in_fpcm; |
45 float* out_fpcm; | 39 float* out_fpcm; |
46 float* noise_fpcm; | 40 float* noise_fpcm; |
47 | |
48 // Current locations in streams | |
49 float* noise_cursor; | 41 float* noise_cursor; |
50 float* clear_cursor; | 42 float* clear_cursor; |
51 | 43 |
52 int samples; // Number of samples in input PCM file | 44 int samples; |
53 int fragment_size; // Number of samples to process at a time | 45 int fragment_size; |
54 // to simulate APM stream processing | |
55 | 46 |
56 using std::complex; | 47 using std::complex; |
57 | |
58 namespace webrtc { | |
59 | |
60 using webrtc::RealFourier; | 48 using webrtc::RealFourier; |
61 using webrtc::IntelligibilityEnhancer; | 49 using webrtc::IntelligibilityEnhancer; |
62 | 50 |
63 DEFINE_int32(clear_type, | 51 DEFINE_int32(clear_type, webrtc::intelligibility::VarianceArray::kStepInfinite, |
64 webrtc::intelligibility::VarianceArray::kStepInfinite, | |
65 "Variance algorithm for clear data."); | 52 "Variance algorithm for clear data."); |
66 DEFINE_double(clear_alpha, 0.9, "Variance decay factor for clear data."); | 53 DEFINE_double(clear_alpha, 0.9, |
67 DEFINE_int32(clear_window, | 54 "Variance decay factor for clear data."); |
68 475, | 55 DEFINE_int32(clear_window, 475, |
69 "Window size for windowed variance for clear data."); | 56 "Window size for windowed variance for clear data."); |
70 DEFINE_int32(sample_rate, | 57 DEFINE_int32(sample_rate, 16000, |
71 16000, | |
72 "Audio sample rate used in the input and output files."); | 58 "Audio sample rate used in the input and output files."); |
73 DEFINE_int32(ana_rate, | 59 DEFINE_int32(ana_rate, 800, |
74 800, | |
75 "Analysis rate; gains recalculated every N blocks."); | 60 "Analysis rate; gains recalculated every N blocks."); |
76 DEFINE_int32( | 61 DEFINE_int32(var_rate, 2, |
77 var_rate, | 62 "Variance clear rate; history is forgotten every N gain recalculati
ons."); |
78 2, | |
79 "Variance clear rate; history is forgotten every N gain recalculations."); | |
80 DEFINE_double(gain_limit, 1000.0, "Maximum gain change in one block."); | 63 DEFINE_double(gain_limit, 1000.0, "Maximum gain change in one block."); |
81 | 64 |
82 DEFINE_bool(repeat, false, "Repeat input file ad nauseam."); | 65 DEFINE_bool(repeat, false, "Repeat input file ad nauseam."); |
83 | 66 |
84 DEFINE_string(clear_file, "speech.pcm", "Input file with clear speech."); | 67 DEFINE_string(clear_file, "speech.pcm", "Input file with clear speech."); |
85 DEFINE_string(noise_file, "noise.pcm", "Input file with noise data."); | 68 DEFINE_string(noise_file, "noise.pcm", "Input file with noise data."); |
86 DEFINE_string(out_file, | 69 DEFINE_string(out_file, "proc_enhanced.pcm", "Enhanced output. Use '-' to " |
87 "proc_enhanced.pcm", | |
88 "Enhanced output. Use '-' to " | |
89 "pipe through aplay internally."); | 70 "pipe through aplay internally."); |
90 | 71 |
91 // Constant IntelligibilityEnhancer constructor parameters. | 72 // Write an Sun AU-formatted audio chunk into file descriptor |fd|. Can be used |
92 const int kErbResolution = 2; | 73 // to pipe the audio stream directly into aplay. |
93 const int kNumChannels = 1; | |
94 | |
95 // Converts output stream to Sun AU format and writes to file descriptor |fd|. | |
96 // Can be used to pipe output directly into aplay. | |
97 // TODO(ekmeyerson): Modify to write WAV instead. | |
98 void writeau(int fd) { | 74 void writeau(int fd) { |
99 uint32_t thing; | 75 uint32_t thing; |
100 | 76 |
101 write(fd, ".snd", 4); | 77 write(fd, ".snd", 4); |
102 thing = htonl(24); | 78 thing = htonl(24); |
103 write(fd, &thing, sizeof(thing)); | 79 write(fd, &thing, sizeof(thing)); |
104 thing = htonl(0xffffffff); | 80 thing = htonl(0xffffffff); |
105 write(fd, &thing, sizeof(thing)); | 81 write(fd, &thing, sizeof(thing)); |
106 thing = htonl(3); | 82 thing = htonl(3); |
107 write(fd, &thing, sizeof(thing)); | 83 write(fd, &thing, sizeof(thing)); |
108 thing = htonl(FLAGS_sample_rate); | 84 thing = htonl(FLAGS_sample_rate); |
109 write(fd, &thing, sizeof(thing)); | 85 write(fd, &thing, sizeof(thing)); |
110 thing = htonl(1); | 86 thing = htonl(1); |
111 write(fd, &thing, sizeof(thing)); | 87 write(fd, &thing, sizeof(thing)); |
112 | 88 |
113 for (int i = 0; i < samples; ++i) { | 89 for (int i = 0; i < samples; ++i) { |
114 out_ipcm[i] = htons(out_ipcm[i]); | 90 out_ipcm[i] = htons(out_ipcm[i]); |
115 } | 91 } |
116 write(fd, out_ipcm, sizeof(*out_ipcm) * samples); | 92 write(fd, out_ipcm, sizeof(*out_ipcm) * samples); |
117 } | 93 } |
118 | 94 |
119 // void function for gtest | 95 int main(int argc, char* argv[]) { |
120 void void_main(int argc, char* argv[]) { | 96 google::SetUsageMessage("\n\nVariance algorithm types are:\n" |
121 google::SetUsageMessage( | 97 " 0 - infinite/normal,\n" |
122 "\n\nVariance algorithm types are:\n" | 98 " 1 - exponentially decaying,\n" |
123 " 0 - infinite/normal,\n" | 99 " 2 - rolling window.\n" |
124 " 1 - exponentially decaying,\n" | 100 "\nInput files must be little-endian 16-bit signed raw
PCM.\n"); |
125 " 2 - rolling window.\n" | |
126 "\nInput files must be little-endian 16-bit signed raw PCM.\n"); | |
127 google::ParseCommandLineFlags(&argc, &argv, true); | 101 google::ParseCommandLineFlags(&argc, &argv, true); |
128 | 102 |
129 const char* in_name = FLAGS_clear_file.c_str(); | 103 const char* in_name = FLAGS_clear_file.c_str(); |
130 const char* out_name = FLAGS_out_file.c_str(); | 104 const char* out_name = FLAGS_out_file.c_str(); |
131 const char* noise_name = FLAGS_noise_file.c_str(); | 105 const char* noise_name = FLAGS_noise_file.c_str(); |
132 struct stat in_stat, noise_stat; | 106 struct stat in_stat, noise_stat; |
133 int in_fd, out_fd, noise_fd; | 107 int in_fd, out_fd, noise_fd; |
134 FILE* aplay_file = nullptr; | 108 FILE* aplay_file = nullptr; |
135 | 109 |
136 // Load settings and set up PCMs. | 110 fragment_size = FLAGS_sample_rate / 100; |
137 | 111 |
138 fragment_size = FLAGS_sample_rate / 100; // Mirror real time APM chunk size. | 112 stat(in_name, &in_stat); |
139 // Duplicates chunk_length_ in | 113 stat(noise_name, &noise_stat); |
140 // IntelligibilityEnhancer. | |
141 | |
142 ASSERT_EQ(stat(in_name, &in_stat), 0) << "Empty speech input."; | |
143 ASSERT_EQ(stat(noise_name, &noise_stat), 0) << "Empty noise input."; | |
144 | |
145 samples = in_stat.st_size / sizeof(*in_ipcm); | 114 samples = in_stat.st_size / sizeof(*in_ipcm); |
146 | 115 |
147 in_fd = open(in_name, O_RDONLY); | 116 in_fd = open(in_name, O_RDONLY); |
148 if (!strcmp(out_name, "-")) { | 117 if (!strcmp(out_name, "-")) { |
149 aplay_file = popen("aplay -t au", "w"); | 118 aplay_file = popen("aplay -t au", "w"); |
150 out_fd = fileno(aplay_file); | 119 out_fd = fileno(aplay_file); |
151 } else { | 120 } else { |
152 out_fd = open(out_name, O_WRONLY | O_CREAT | O_TRUNC, | 121 out_fd = open(out_name, O_WRONLY | O_CREAT | O_TRUNC, |
153 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH); | 122 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH | S_IWOTH); |
154 } | 123 } |
155 noise_fd = open(noise_name, O_RDONLY); | 124 noise_fd = open(noise_name, O_RDONLY); |
156 | 125 |
157 in_ipcm = static_cast<int16_t*>( | 126 in_ipcm = static_cast<int16_t*>(mmap(nullptr, in_stat.st_size, PROT_READ, |
158 mmap(nullptr, in_stat.st_size, PROT_READ, MAP_PRIVATE, in_fd, 0)); | 127 MAP_PRIVATE, in_fd, 0)); |
159 noise_ipcm = static_cast<int16_t*>( | 128 noise_ipcm = static_cast<int16_t*>(mmap(nullptr, noise_stat.st_size, |
160 mmap(nullptr, noise_stat.st_size, PROT_READ, MAP_PRIVATE, noise_fd, 0)); | 129 PROT_READ, MAP_PRIVATE, noise_fd, 0)); |
161 out_ipcm = new int16_t[samples]; | 130 out_ipcm = new int16_t[samples]; |
162 out_fpcm = new float[samples]; | 131 out_fpcm = new float[samples]; |
163 in_fpcm = new float[samples]; | 132 in_fpcm = new float[samples]; |
164 noise_fpcm = new float[samples]; | 133 noise_fpcm = new float[samples]; |
165 | 134 |
166 for (int i = 0; i < samples; ++i) { | 135 for (int i = 0; i < samples; ++i) { |
167 noise_fpcm[i] = noise_ipcm[i % (noise_stat.st_size / sizeof(*noise_ipcm))]; | 136 noise_fpcm[i] = noise_ipcm[i % (noise_stat.st_size / sizeof(*noise_ipcm))]; |
168 } | 137 } |
169 | 138 |
170 // Run intelligibility enhancement. | 139 //feenableexcept(FE_INVALID | FE_OVERFLOW); |
171 | 140 IntelligibilityEnhancer enh(2, |
172 IntelligibilityEnhancer enh( | 141 FLAGS_sample_rate, 1, |
173 kErbResolution, | 142 FLAGS_clear_type, |
174 FLAGS_sample_rate, | 143 static_cast<float>(FLAGS_clear_alpha), |
175 kNumChannels, | 144 FLAGS_clear_window, |
176 FLAGS_clear_type, static_cast<float>(FLAGS_clear_alpha), | 145 FLAGS_ana_rate, |
177 FLAGS_clear_window, FLAGS_ana_rate, FLAGS_var_rate, FLAGS_gain_limit); | 146 FLAGS_var_rate, |
| 147 FLAGS_gain_limit); |
178 | 148 |
179 // Slice the input into smaller chunks, as the APM would do, and feed them | 149 // Slice the input into smaller chunks, as the APM would do, and feed them |
180 // through the enhancer. Repeat indefinitely if FLAGS_repeat is set. | 150 // into the enhancer. Repeat indefinitely if FLAGS_repeat is set. |
181 do { | 151 do { |
182 noise_cursor = noise_fpcm; | 152 noise_cursor = noise_fpcm; |
183 clear_cursor = in_fpcm; | 153 clear_cursor = in_fpcm; |
184 for (int i = 0; i < samples; ++i) { | 154 for (int i = 0; i < samples; ++i) { |
185 in_fpcm[i] = in_ipcm[i]; | 155 in_fpcm[i] = in_ipcm[i]; |
186 } | 156 } |
187 | 157 |
188 for (int i = 0; i < samples; i += fragment_size) { | 158 for (int i = 0; i < samples; i += fragment_size) { |
189 enh.ProcessCaptureAudio(&noise_cursor); | 159 enh.ProcessCaptureAudio(&noise_cursor); |
190 enh.ProcessRenderAudio(&clear_cursor); | 160 enh.ProcessRenderAudio(&clear_cursor); |
(...skipping 13 matching lines...) Expand all Loading... |
204 | 174 |
205 munmap(const_cast<int16_t*>(noise_ipcm), noise_stat.st_size); | 175 munmap(const_cast<int16_t*>(noise_ipcm), noise_stat.st_size); |
206 munmap(const_cast<int16_t*>(in_ipcm), in_stat.st_size); | 176 munmap(const_cast<int16_t*>(in_ipcm), in_stat.st_size); |
207 close(noise_fd); | 177 close(noise_fd); |
208 if (aplay_file) { | 178 if (aplay_file) { |
209 pclose(aplay_file); | 179 pclose(aplay_file); |
210 } else { | 180 } else { |
211 close(out_fd); | 181 close(out_fd); |
212 } | 182 } |
213 close(in_fd); | 183 close(in_fd); |
| 184 |
| 185 return 0; |
214 } | 186 } |
215 | 187 |
216 } // namespace webrtc | |
217 | |
218 int main(int argc, char* argv[]) { | |
219 webrtc::void_main(argc, argv); | |
220 return 0; | |
221 } | |
OLD | NEW |