andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2012 The WebRTC project authors. All Rights Reserved. |
| 3 | * |
| 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 |
| 6 | * tree. An additional intellectual property rights grant can be found |
| 7 | * in the file PATENTS. All contributing project authors may |
| 8 | * be found in the AUTHORS file in the root of the source tree. |
| 9 | */ |
| 10 | |
andrew@webrtc.org | f1de5e9 | 2013-03-01 16:36:19 +0000 | [diff] [blame] | 11 | #include "webrtc/modules/audio_processing/audio_processing_impl.h" |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 12 | |
| 13 | #include <assert.h> |
| 14 | |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 15 | #include "webrtc/common_audio/signal_processing/include/signal_processing_library.h" |
andrew@webrtc.org | f1de5e9 | 2013-03-01 16:36:19 +0000 | [diff] [blame] | 16 | #include "webrtc/modules/audio_processing/audio_buffer.h" |
andrew@webrtc.org | 25613ea | 2013-07-25 18:28:29 +0000 | [diff] [blame] | 17 | #include "webrtc/modules/audio_processing/echo_cancellation_impl_wrapper.h" |
andrew@webrtc.org | f1de5e9 | 2013-03-01 16:36:19 +0000 | [diff] [blame] | 18 | #include "webrtc/modules/audio_processing/echo_control_mobile_impl.h" |
| 19 | #include "webrtc/modules/audio_processing/gain_control_impl.h" |
| 20 | #include "webrtc/modules/audio_processing/high_pass_filter_impl.h" |
| 21 | #include "webrtc/modules/audio_processing/level_estimator_impl.h" |
| 22 | #include "webrtc/modules/audio_processing/noise_suppression_impl.h" |
| 23 | #include "webrtc/modules/audio_processing/processing_component.h" |
andrew@webrtc.org | f1de5e9 | 2013-03-01 16:36:19 +0000 | [diff] [blame] | 24 | #include "webrtc/modules/audio_processing/voice_detection_impl.h" |
| 25 | #include "webrtc/modules/interface/module_common_types.h" |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 26 | #include "webrtc/system_wrappers/interface/compile_assert.h" |
andrew@webrtc.org | f1de5e9 | 2013-03-01 16:36:19 +0000 | [diff] [blame] | 27 | #include "webrtc/system_wrappers/interface/critical_section_wrapper.h" |
| 28 | #include "webrtc/system_wrappers/interface/file_wrapper.h" |
| 29 | #include "webrtc/system_wrappers/interface/logging.h" |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 30 | |
| 31 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 32 | // Files generated at build-time by the protobuf compiler. |
| 33 | #ifdef WEBRTC_ANDROID_PLATFORM_BUILD |
leozwang@webrtc.org | 5e512ae | 2012-10-22 21:21:52 +0000 | [diff] [blame] | 34 | #include "external/webrtc/webrtc/modules/audio_processing/debug.pb.h" |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 35 | #else |
| 36 | #include "webrtc/audio_processing/debug.pb.h" |
| 37 | #endif |
| 38 | #endif // WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 39 | |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 40 | static const int kChunkSizeMs = 10; |
| 41 | |
| 42 | #define RETURN_ON_ERR(expr) \ |
| 43 | do { \ |
| 44 | int err = expr; \ |
| 45 | if (err != kNoError) { \ |
| 46 | return err; \ |
| 47 | } \ |
| 48 | } while (0) |
| 49 | |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 50 | namespace webrtc { |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 51 | |
| 52 | // Throughout webrtc, it's assumed that success is represented by zero. |
| 53 | COMPILE_ASSERT(AudioProcessing::kNoError == 0, no_error_must_be_zero); |
| 54 | |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 55 | AudioProcessing* AudioProcessing::Create(int id) { |
andrew@webrtc.org | 3fbe666 | 2014-01-25 02:09:06 +0000 | [diff] [blame] | 56 | return Create(); |
| 57 | } |
| 58 | |
| 59 | AudioProcessing* AudioProcessing::Create() { |
| 60 | Config config; |
| 61 | return Create(config); |
| 62 | } |
| 63 | |
| 64 | AudioProcessing* AudioProcessing::Create(const Config& config) { |
| 65 | AudioProcessingImpl* apm = new AudioProcessingImpl(config); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 66 | if (apm->Initialize() != kNoError) { |
| 67 | delete apm; |
| 68 | apm = NULL; |
| 69 | } |
| 70 | |
| 71 | return apm; |
| 72 | } |
| 73 | |
pbos@webrtc.org | 24add92 | 2013-08-02 11:44:11 +0000 | [diff] [blame] | 74 | int32_t AudioProcessing::TimeUntilNextProcess() { return -1; } |
| 75 | int32_t AudioProcessing::Process() { return -1; } |
| 76 | |
andrew@webrtc.org | 3fbe666 | 2014-01-25 02:09:06 +0000 | [diff] [blame] | 77 | AudioProcessingImpl::AudioProcessingImpl(const Config& config) |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 78 | : echo_cancellation_(NULL), |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 79 | echo_control_mobile_(NULL), |
| 80 | gain_control_(NULL), |
| 81 | high_pass_filter_(NULL), |
| 82 | level_estimator_(NULL), |
| 83 | noise_suppression_(NULL), |
| 84 | voice_detection_(NULL), |
| 85 | crit_(CriticalSectionWrapper::CreateCriticalSection()), |
| 86 | render_audio_(NULL), |
| 87 | capture_audio_(NULL), |
| 88 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 89 | debug_file_(FileWrapper::Create()), |
| 90 | event_msg_(new audioproc::Event()), |
| 91 | #endif |
| 92 | sample_rate_hz_(kSampleRate16kHz), |
| 93 | split_sample_rate_hz_(kSampleRate16kHz), |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 94 | samples_per_channel_(kChunkSizeMs * sample_rate_hz_ / 1000), |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 95 | stream_delay_ms_(0), |
| 96 | delay_offset_ms_(0), |
| 97 | was_stream_delay_set_(false), |
| 98 | num_reverse_channels_(1), |
| 99 | num_input_channels_(1), |
| 100 | num_output_channels_(1) { |
andrew@webrtc.org | 25613ea | 2013-07-25 18:28:29 +0000 | [diff] [blame] | 101 | echo_cancellation_ = EchoCancellationImplWrapper::Create(this); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 102 | component_list_.push_back(echo_cancellation_); |
| 103 | |
| 104 | echo_control_mobile_ = new EchoControlMobileImpl(this); |
| 105 | component_list_.push_back(echo_control_mobile_); |
| 106 | |
| 107 | gain_control_ = new GainControlImpl(this); |
| 108 | component_list_.push_back(gain_control_); |
| 109 | |
| 110 | high_pass_filter_ = new HighPassFilterImpl(this); |
| 111 | component_list_.push_back(high_pass_filter_); |
| 112 | |
| 113 | level_estimator_ = new LevelEstimatorImpl(this); |
| 114 | component_list_.push_back(level_estimator_); |
| 115 | |
| 116 | noise_suppression_ = new NoiseSuppressionImpl(this); |
| 117 | component_list_.push_back(noise_suppression_); |
| 118 | |
| 119 | voice_detection_ = new VoiceDetectionImpl(this); |
| 120 | component_list_.push_back(voice_detection_); |
andrew@webrtc.org | 3fbe666 | 2014-01-25 02:09:06 +0000 | [diff] [blame] | 121 | |
| 122 | SetExtraOptions(config); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 123 | } |
| 124 | |
| 125 | AudioProcessingImpl::~AudioProcessingImpl() { |
andrew@webrtc.org | 7838d79 | 2012-10-27 00:28:27 +0000 | [diff] [blame] | 126 | { |
| 127 | CriticalSectionScoped crit_scoped(crit_); |
| 128 | while (!component_list_.empty()) { |
| 129 | ProcessingComponent* component = component_list_.front(); |
| 130 | component->Destroy(); |
| 131 | delete component; |
| 132 | component_list_.pop_front(); |
| 133 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 134 | |
| 135 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
andrew@webrtc.org | 7838d79 | 2012-10-27 00:28:27 +0000 | [diff] [blame] | 136 | if (debug_file_->Open()) { |
| 137 | debug_file_->CloseFile(); |
| 138 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 139 | #endif |
| 140 | |
andrew@webrtc.org | 7838d79 | 2012-10-27 00:28:27 +0000 | [diff] [blame] | 141 | if (render_audio_) { |
| 142 | delete render_audio_; |
| 143 | render_audio_ = NULL; |
| 144 | } |
| 145 | |
| 146 | if (capture_audio_) { |
| 147 | delete capture_audio_; |
| 148 | capture_audio_ = NULL; |
| 149 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 150 | } |
| 151 | |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 152 | delete crit_; |
| 153 | crit_ = NULL; |
| 154 | } |
| 155 | |
| 156 | CriticalSectionWrapper* AudioProcessingImpl::crit() const { |
| 157 | return crit_; |
| 158 | } |
| 159 | |
| 160 | int AudioProcessingImpl::split_sample_rate_hz() const { |
| 161 | return split_sample_rate_hz_; |
| 162 | } |
| 163 | |
| 164 | int AudioProcessingImpl::Initialize() { |
| 165 | CriticalSectionScoped crit_scoped(crit_); |
| 166 | return InitializeLocked(); |
| 167 | } |
| 168 | |
| 169 | int AudioProcessingImpl::InitializeLocked() { |
| 170 | if (render_audio_ != NULL) { |
| 171 | delete render_audio_; |
| 172 | render_audio_ = NULL; |
| 173 | } |
| 174 | |
| 175 | if (capture_audio_ != NULL) { |
| 176 | delete capture_audio_; |
| 177 | capture_audio_ = NULL; |
| 178 | } |
| 179 | |
| 180 | render_audio_ = new AudioBuffer(num_reverse_channels_, |
| 181 | samples_per_channel_); |
| 182 | capture_audio_ = new AudioBuffer(num_input_channels_, |
| 183 | samples_per_channel_); |
| 184 | |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 185 | // Initialize all components. |
| 186 | std::list<ProcessingComponent*>::iterator it; |
andrew@webrtc.org | 7838d79 | 2012-10-27 00:28:27 +0000 | [diff] [blame] | 187 | for (it = component_list_.begin(); it != component_list_.end(); ++it) { |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 188 | int err = (*it)->Initialize(); |
| 189 | if (err != kNoError) { |
| 190 | return err; |
| 191 | } |
| 192 | } |
| 193 | |
| 194 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 195 | if (debug_file_->Open()) { |
| 196 | int err = WriteInitMessage(); |
| 197 | if (err != kNoError) { |
| 198 | return err; |
| 199 | } |
| 200 | } |
| 201 | #endif |
| 202 | |
| 203 | return kNoError; |
| 204 | } |
| 205 | |
andrew@webrtc.org | 25613ea | 2013-07-25 18:28:29 +0000 | [diff] [blame] | 206 | void AudioProcessingImpl::SetExtraOptions(const Config& config) { |
andrew@webrtc.org | 3fbe666 | 2014-01-25 02:09:06 +0000 | [diff] [blame] | 207 | CriticalSectionScoped crit_scoped(crit_); |
andrew@webrtc.org | 25613ea | 2013-07-25 18:28:29 +0000 | [diff] [blame] | 208 | std::list<ProcessingComponent*>::iterator it; |
| 209 | for (it = component_list_.begin(); it != component_list_.end(); ++it) |
| 210 | (*it)->SetExtraOptions(config); |
| 211 | } |
| 212 | |
aluebs@webrtc.org | 09b40ec | 2013-11-19 15:17:51 +0000 | [diff] [blame] | 213 | int AudioProcessingImpl::EnableExperimentalNs(bool enable) { |
| 214 | return kNoError; |
| 215 | } |
| 216 | |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 217 | int AudioProcessingImpl::set_sample_rate_hz(int rate) { |
| 218 | CriticalSectionScoped crit_scoped(crit_); |
andrew@webrtc.org | 7838d79 | 2012-10-27 00:28:27 +0000 | [diff] [blame] | 219 | if (rate == sample_rate_hz_) { |
| 220 | return kNoError; |
| 221 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 222 | if (rate != kSampleRate8kHz && |
| 223 | rate != kSampleRate16kHz && |
| 224 | rate != kSampleRate32kHz) { |
| 225 | return kBadParameterError; |
| 226 | } |
andrew@webrtc.org | f1de5e9 | 2013-03-01 16:36:19 +0000 | [diff] [blame] | 227 | if (echo_control_mobile_->is_enabled() && rate > kSampleRate16kHz) { |
| 228 | LOG(LS_ERROR) << "AECM only supports 16 kHz or lower sample rates"; |
| 229 | return kUnsupportedComponentError; |
| 230 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 231 | |
| 232 | sample_rate_hz_ = rate; |
| 233 | samples_per_channel_ = rate / 100; |
| 234 | |
| 235 | if (sample_rate_hz_ == kSampleRate32kHz) { |
| 236 | split_sample_rate_hz_ = kSampleRate16kHz; |
| 237 | } else { |
| 238 | split_sample_rate_hz_ = sample_rate_hz_; |
| 239 | } |
| 240 | |
| 241 | return InitializeLocked(); |
| 242 | } |
| 243 | |
| 244 | int AudioProcessingImpl::sample_rate_hz() const { |
henrika@webrtc.org | 1d25eac | 2013-04-05 14:34:57 +0000 | [diff] [blame] | 245 | CriticalSectionScoped crit_scoped(crit_); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 246 | return sample_rate_hz_; |
| 247 | } |
| 248 | |
| 249 | int AudioProcessingImpl::set_num_reverse_channels(int channels) { |
| 250 | CriticalSectionScoped crit_scoped(crit_); |
andrew@webrtc.org | 7838d79 | 2012-10-27 00:28:27 +0000 | [diff] [blame] | 251 | if (channels == num_reverse_channels_) { |
| 252 | return kNoError; |
| 253 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 254 | // Only stereo supported currently. |
| 255 | if (channels > 2 || channels < 1) { |
| 256 | return kBadParameterError; |
| 257 | } |
| 258 | |
| 259 | num_reverse_channels_ = channels; |
| 260 | |
| 261 | return InitializeLocked(); |
| 262 | } |
| 263 | |
| 264 | int AudioProcessingImpl::num_reverse_channels() const { |
| 265 | return num_reverse_channels_; |
| 266 | } |
| 267 | |
| 268 | int AudioProcessingImpl::set_num_channels( |
| 269 | int input_channels, |
| 270 | int output_channels) { |
| 271 | CriticalSectionScoped crit_scoped(crit_); |
andrew@webrtc.org | 7838d79 | 2012-10-27 00:28:27 +0000 | [diff] [blame] | 272 | if (input_channels == num_input_channels_ && |
| 273 | output_channels == num_output_channels_) { |
| 274 | return kNoError; |
| 275 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 276 | if (output_channels > input_channels) { |
| 277 | return kBadParameterError; |
| 278 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 279 | // Only stereo supported currently. |
andrew@webrtc.org | 7838d79 | 2012-10-27 00:28:27 +0000 | [diff] [blame] | 280 | if (input_channels > 2 || input_channels < 1 || |
| 281 | output_channels > 2 || output_channels < 1) { |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 282 | return kBadParameterError; |
| 283 | } |
| 284 | |
| 285 | num_input_channels_ = input_channels; |
| 286 | num_output_channels_ = output_channels; |
| 287 | |
| 288 | return InitializeLocked(); |
| 289 | } |
| 290 | |
| 291 | int AudioProcessingImpl::num_input_channels() const { |
| 292 | return num_input_channels_; |
| 293 | } |
| 294 | |
| 295 | int AudioProcessingImpl::num_output_channels() const { |
| 296 | return num_output_channels_; |
| 297 | } |
| 298 | |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 299 | int AudioProcessingImpl::MaybeInitializeLocked(int sample_rate_hz, |
| 300 | int num_input_channels, int num_output_channels, int num_reverse_channels) { |
| 301 | if (sample_rate_hz == sample_rate_hz_ && |
| 302 | num_input_channels == num_input_channels_ && |
| 303 | num_output_channels == num_output_channels_ && |
| 304 | num_reverse_channels == num_reverse_channels_) { |
| 305 | return kNoError; |
| 306 | } |
| 307 | |
| 308 | if (sample_rate_hz != kSampleRate8kHz && |
| 309 | sample_rate_hz != kSampleRate16kHz && |
| 310 | sample_rate_hz != kSampleRate32kHz) { |
| 311 | return kBadSampleRateError; |
| 312 | } |
| 313 | if (num_output_channels > num_input_channels) { |
| 314 | return kBadNumberChannelsError; |
| 315 | } |
| 316 | // Only mono and stereo supported currently. |
| 317 | if (num_input_channels > 2 || num_input_channels < 1 || |
| 318 | num_output_channels > 2 || num_output_channels < 1 || |
| 319 | num_reverse_channels > 2 || num_reverse_channels < 1) { |
| 320 | return kBadNumberChannelsError; |
| 321 | } |
| 322 | if (echo_control_mobile_->is_enabled() && sample_rate_hz > kSampleRate16kHz) { |
| 323 | LOG(LS_ERROR) << "AECM only supports 16 or 8 kHz sample rates"; |
| 324 | return kUnsupportedComponentError; |
| 325 | } |
| 326 | |
| 327 | sample_rate_hz_ = sample_rate_hz; |
| 328 | samples_per_channel_ = kChunkSizeMs * sample_rate_hz / 1000; |
| 329 | num_input_channels_ = num_input_channels; |
| 330 | num_output_channels_ = num_output_channels; |
| 331 | num_reverse_channels_ = num_reverse_channels; |
| 332 | |
| 333 | if (sample_rate_hz_ == kSampleRate32kHz) { |
| 334 | split_sample_rate_hz_ = kSampleRate16kHz; |
| 335 | } else { |
| 336 | split_sample_rate_hz_ = sample_rate_hz_; |
| 337 | } |
| 338 | |
| 339 | return InitializeLocked(); |
| 340 | } |
| 341 | |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 342 | int AudioProcessingImpl::ProcessStream(AudioFrame* frame) { |
| 343 | CriticalSectionScoped crit_scoped(crit_); |
| 344 | int err = kNoError; |
| 345 | |
| 346 | if (frame == NULL) { |
| 347 | return kNullPointerError; |
| 348 | } |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 349 | // TODO(ajm): We now always set the output channels equal to the input |
| 350 | // channels here. Remove the ability to downmix entirely. |
| 351 | RETURN_ON_ERR(MaybeInitializeLocked(frame->sample_rate_hz_, |
| 352 | frame->num_channels_, frame->num_channels_, num_reverse_channels_)); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 353 | if (frame->samples_per_channel_ != samples_per_channel_) { |
| 354 | return kBadDataLengthError; |
| 355 | } |
| 356 | |
| 357 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 358 | if (debug_file_->Open()) { |
| 359 | event_msg_->set_type(audioproc::Event::STREAM); |
| 360 | audioproc::Stream* msg = event_msg_->mutable_stream(); |
| 361 | const size_t data_size = sizeof(int16_t) * |
| 362 | frame->samples_per_channel_ * |
| 363 | frame->num_channels_; |
| 364 | msg->set_input_data(frame->data_, data_size); |
| 365 | msg->set_delay(stream_delay_ms_); |
| 366 | msg->set_drift(echo_cancellation_->stream_drift_samples()); |
| 367 | msg->set_level(gain_control_->stream_analog_level()); |
| 368 | } |
| 369 | #endif |
| 370 | |
| 371 | capture_audio_->DeinterleaveFrom(frame); |
| 372 | |
| 373 | // TODO(ajm): experiment with mixing and AEC placement. |
| 374 | if (num_output_channels_ < num_input_channels_) { |
| 375 | capture_audio_->Mix(num_output_channels_); |
| 376 | frame->num_channels_ = num_output_channels_; |
| 377 | } |
| 378 | |
| 379 | bool data_processed = is_data_processed(); |
| 380 | if (analysis_needed(data_processed)) { |
| 381 | for (int i = 0; i < num_output_channels_; i++) { |
| 382 | // Split into a low and high band. |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 383 | WebRtcSpl_AnalysisQMF(capture_audio_->data(i), |
| 384 | capture_audio_->samples_per_channel(), |
| 385 | capture_audio_->low_pass_split_data(i), |
| 386 | capture_audio_->high_pass_split_data(i), |
| 387 | capture_audio_->analysis_filter_state1(i), |
| 388 | capture_audio_->analysis_filter_state2(i)); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 389 | } |
| 390 | } |
| 391 | |
| 392 | err = high_pass_filter_->ProcessCaptureAudio(capture_audio_); |
| 393 | if (err != kNoError) { |
| 394 | return err; |
| 395 | } |
| 396 | |
| 397 | err = gain_control_->AnalyzeCaptureAudio(capture_audio_); |
| 398 | if (err != kNoError) { |
| 399 | return err; |
| 400 | } |
| 401 | |
| 402 | err = echo_cancellation_->ProcessCaptureAudio(capture_audio_); |
| 403 | if (err != kNoError) { |
| 404 | return err; |
| 405 | } |
| 406 | |
| 407 | if (echo_control_mobile_->is_enabled() && |
| 408 | noise_suppression_->is_enabled()) { |
| 409 | capture_audio_->CopyLowPassToReference(); |
| 410 | } |
| 411 | |
| 412 | err = noise_suppression_->ProcessCaptureAudio(capture_audio_); |
| 413 | if (err != kNoError) { |
| 414 | return err; |
| 415 | } |
| 416 | |
| 417 | err = echo_control_mobile_->ProcessCaptureAudio(capture_audio_); |
| 418 | if (err != kNoError) { |
| 419 | return err; |
| 420 | } |
| 421 | |
| 422 | err = voice_detection_->ProcessCaptureAudio(capture_audio_); |
| 423 | if (err != kNoError) { |
| 424 | return err; |
| 425 | } |
| 426 | |
| 427 | err = gain_control_->ProcessCaptureAudio(capture_audio_); |
| 428 | if (err != kNoError) { |
| 429 | return err; |
| 430 | } |
| 431 | |
| 432 | if (synthesis_needed(data_processed)) { |
| 433 | for (int i = 0; i < num_output_channels_; i++) { |
| 434 | // Recombine low and high bands. |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 435 | WebRtcSpl_SynthesisQMF(capture_audio_->low_pass_split_data(i), |
| 436 | capture_audio_->high_pass_split_data(i), |
| 437 | capture_audio_->samples_per_split_channel(), |
| 438 | capture_audio_->data(i), |
| 439 | capture_audio_->synthesis_filter_state1(i), |
| 440 | capture_audio_->synthesis_filter_state2(i)); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 441 | } |
| 442 | } |
| 443 | |
| 444 | // The level estimator operates on the recombined data. |
| 445 | err = level_estimator_->ProcessStream(capture_audio_); |
| 446 | if (err != kNoError) { |
| 447 | return err; |
| 448 | } |
| 449 | |
| 450 | capture_audio_->InterleaveTo(frame, interleave_needed(data_processed)); |
| 451 | |
| 452 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 453 | if (debug_file_->Open()) { |
| 454 | audioproc::Stream* msg = event_msg_->mutable_stream(); |
| 455 | const size_t data_size = sizeof(int16_t) * |
| 456 | frame->samples_per_channel_ * |
| 457 | frame->num_channels_; |
| 458 | msg->set_output_data(frame->data_, data_size); |
| 459 | err = WriteMessageToDebugFile(); |
| 460 | if (err != kNoError) { |
| 461 | return err; |
| 462 | } |
| 463 | } |
| 464 | #endif |
| 465 | |
| 466 | was_stream_delay_set_ = false; |
| 467 | return kNoError; |
| 468 | } |
| 469 | |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 470 | // TODO(ajm): Have AnalyzeReverseStream accept sample rates not matching the |
| 471 | // primary stream and convert ourselves rather than having the user manage it. |
| 472 | // We can be smarter and use the splitting filter when appropriate. Similarly, |
| 473 | // perform downmixing here. |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 474 | int AudioProcessingImpl::AnalyzeReverseStream(AudioFrame* frame) { |
| 475 | CriticalSectionScoped crit_scoped(crit_); |
| 476 | int err = kNoError; |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 477 | if (frame == NULL) { |
| 478 | return kNullPointerError; |
| 479 | } |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 480 | if (frame->sample_rate_hz_ != sample_rate_hz_) { |
| 481 | return kBadSampleRateError; |
| 482 | } |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 483 | RETURN_ON_ERR(MaybeInitializeLocked(sample_rate_hz_, num_input_channels_, |
| 484 | num_output_channels_, frame->num_channels_)); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 485 | |
| 486 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 487 | if (debug_file_->Open()) { |
| 488 | event_msg_->set_type(audioproc::Event::REVERSE_STREAM); |
| 489 | audioproc::ReverseStream* msg = event_msg_->mutable_reverse_stream(); |
| 490 | const size_t data_size = sizeof(int16_t) * |
| 491 | frame->samples_per_channel_ * |
| 492 | frame->num_channels_; |
| 493 | msg->set_data(frame->data_, data_size); |
| 494 | err = WriteMessageToDebugFile(); |
| 495 | if (err != kNoError) { |
| 496 | return err; |
| 497 | } |
| 498 | } |
| 499 | #endif |
| 500 | |
| 501 | render_audio_->DeinterleaveFrom(frame); |
| 502 | |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 503 | if (sample_rate_hz_ == kSampleRate32kHz) { |
| 504 | for (int i = 0; i < num_reverse_channels_; i++) { |
| 505 | // Split into low and high band. |
andrew@webrtc.org | e95dc25 | 2014-01-07 17:45:09 +0000 | [diff] [blame] | 506 | WebRtcSpl_AnalysisQMF(render_audio_->data(i), |
| 507 | render_audio_->samples_per_channel(), |
| 508 | render_audio_->low_pass_split_data(i), |
| 509 | render_audio_->high_pass_split_data(i), |
| 510 | render_audio_->analysis_filter_state1(i), |
| 511 | render_audio_->analysis_filter_state2(i)); |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 512 | } |
| 513 | } |
| 514 | |
| 515 | // TODO(ajm): warnings possible from components? |
| 516 | err = echo_cancellation_->ProcessRenderAudio(render_audio_); |
| 517 | if (err != kNoError) { |
| 518 | return err; |
| 519 | } |
| 520 | |
| 521 | err = echo_control_mobile_->ProcessRenderAudio(render_audio_); |
| 522 | if (err != kNoError) { |
| 523 | return err; |
| 524 | } |
| 525 | |
| 526 | err = gain_control_->ProcessRenderAudio(render_audio_); |
| 527 | if (err != kNoError) { |
| 528 | return err; |
| 529 | } |
| 530 | |
| 531 | return err; // TODO(ajm): this is for returning warnings; necessary? |
| 532 | } |
| 533 | |
| 534 | int AudioProcessingImpl::set_stream_delay_ms(int delay) { |
| 535 | Error retval = kNoError; |
| 536 | was_stream_delay_set_ = true; |
| 537 | delay += delay_offset_ms_; |
| 538 | |
| 539 | if (delay < 0) { |
| 540 | delay = 0; |
| 541 | retval = kBadStreamParameterWarning; |
| 542 | } |
| 543 | |
| 544 | // TODO(ajm): the max is rather arbitrarily chosen; investigate. |
| 545 | if (delay > 500) { |
| 546 | delay = 500; |
| 547 | retval = kBadStreamParameterWarning; |
| 548 | } |
| 549 | |
| 550 | stream_delay_ms_ = delay; |
| 551 | return retval; |
| 552 | } |
| 553 | |
| 554 | int AudioProcessingImpl::stream_delay_ms() const { |
| 555 | return stream_delay_ms_; |
| 556 | } |
| 557 | |
| 558 | bool AudioProcessingImpl::was_stream_delay_set() const { |
| 559 | return was_stream_delay_set_; |
| 560 | } |
| 561 | |
| 562 | void AudioProcessingImpl::set_delay_offset_ms(int offset) { |
| 563 | CriticalSectionScoped crit_scoped(crit_); |
| 564 | delay_offset_ms_ = offset; |
| 565 | } |
| 566 | |
| 567 | int AudioProcessingImpl::delay_offset_ms() const { |
| 568 | return delay_offset_ms_; |
| 569 | } |
| 570 | |
| 571 | int AudioProcessingImpl::StartDebugRecording( |
| 572 | const char filename[AudioProcessing::kMaxFilenameSize]) { |
| 573 | CriticalSectionScoped crit_scoped(crit_); |
| 574 | assert(kMaxFilenameSize == FileWrapper::kMaxFileNameSize); |
| 575 | |
| 576 | if (filename == NULL) { |
| 577 | return kNullPointerError; |
| 578 | } |
| 579 | |
| 580 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 581 | // Stop any ongoing recording. |
| 582 | if (debug_file_->Open()) { |
| 583 | if (debug_file_->CloseFile() == -1) { |
| 584 | return kFileError; |
| 585 | } |
| 586 | } |
| 587 | |
| 588 | if (debug_file_->OpenFile(filename, false) == -1) { |
| 589 | debug_file_->CloseFile(); |
| 590 | return kFileError; |
| 591 | } |
| 592 | |
| 593 | int err = WriteInitMessage(); |
| 594 | if (err != kNoError) { |
| 595 | return err; |
| 596 | } |
| 597 | return kNoError; |
| 598 | #else |
| 599 | return kUnsupportedFunctionError; |
| 600 | #endif // WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 601 | } |
| 602 | |
henrikg@webrtc.org | 7b72264 | 2013-12-06 16:05:17 +0000 | [diff] [blame] | 603 | int AudioProcessingImpl::StartDebugRecording(FILE* handle) { |
| 604 | CriticalSectionScoped crit_scoped(crit_); |
| 605 | |
| 606 | if (handle == NULL) { |
| 607 | return kNullPointerError; |
| 608 | } |
| 609 | |
| 610 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 611 | // Stop any ongoing recording. |
| 612 | if (debug_file_->Open()) { |
| 613 | if (debug_file_->CloseFile() == -1) { |
| 614 | return kFileError; |
| 615 | } |
| 616 | } |
| 617 | |
| 618 | if (debug_file_->OpenFromFileHandle(handle, true, false) == -1) { |
| 619 | return kFileError; |
| 620 | } |
| 621 | |
| 622 | int err = WriteInitMessage(); |
| 623 | if (err != kNoError) { |
| 624 | return err; |
| 625 | } |
| 626 | return kNoError; |
| 627 | #else |
| 628 | return kUnsupportedFunctionError; |
| 629 | #endif // WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 630 | } |
| 631 | |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 632 | int AudioProcessingImpl::StopDebugRecording() { |
| 633 | CriticalSectionScoped crit_scoped(crit_); |
| 634 | |
| 635 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 636 | // We just return if recording hasn't started. |
| 637 | if (debug_file_->Open()) { |
| 638 | if (debug_file_->CloseFile() == -1) { |
| 639 | return kFileError; |
| 640 | } |
| 641 | } |
| 642 | return kNoError; |
| 643 | #else |
| 644 | return kUnsupportedFunctionError; |
| 645 | #endif // WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 646 | } |
| 647 | |
| 648 | EchoCancellation* AudioProcessingImpl::echo_cancellation() const { |
| 649 | return echo_cancellation_; |
| 650 | } |
| 651 | |
| 652 | EchoControlMobile* AudioProcessingImpl::echo_control_mobile() const { |
| 653 | return echo_control_mobile_; |
| 654 | } |
| 655 | |
| 656 | GainControl* AudioProcessingImpl::gain_control() const { |
| 657 | return gain_control_; |
| 658 | } |
| 659 | |
| 660 | HighPassFilter* AudioProcessingImpl::high_pass_filter() const { |
| 661 | return high_pass_filter_; |
| 662 | } |
| 663 | |
| 664 | LevelEstimator* AudioProcessingImpl::level_estimator() const { |
| 665 | return level_estimator_; |
| 666 | } |
| 667 | |
| 668 | NoiseSuppression* AudioProcessingImpl::noise_suppression() const { |
| 669 | return noise_suppression_; |
| 670 | } |
| 671 | |
| 672 | VoiceDetection* AudioProcessingImpl::voice_detection() const { |
| 673 | return voice_detection_; |
| 674 | } |
| 675 | |
pbos@webrtc.org | 3f6d5e0 | 2013-04-10 07:50:54 +0000 | [diff] [blame] | 676 | int32_t AudioProcessingImpl::ChangeUniqueId(const int32_t id) { |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 677 | return kNoError; |
| 678 | } |
| 679 | |
| 680 | bool AudioProcessingImpl::is_data_processed() const { |
| 681 | int enabled_count = 0; |
| 682 | std::list<ProcessingComponent*>::const_iterator it; |
| 683 | for (it = component_list_.begin(); it != component_list_.end(); it++) { |
| 684 | if ((*it)->is_component_enabled()) { |
| 685 | enabled_count++; |
| 686 | } |
| 687 | } |
| 688 | |
| 689 | // Data is unchanged if no components are enabled, or if only level_estimator_ |
| 690 | // or voice_detection_ is enabled. |
| 691 | if (enabled_count == 0) { |
| 692 | return false; |
| 693 | } else if (enabled_count == 1) { |
| 694 | if (level_estimator_->is_enabled() || voice_detection_->is_enabled()) { |
| 695 | return false; |
| 696 | } |
| 697 | } else if (enabled_count == 2) { |
| 698 | if (level_estimator_->is_enabled() && voice_detection_->is_enabled()) { |
| 699 | return false; |
| 700 | } |
| 701 | } |
| 702 | return true; |
| 703 | } |
| 704 | |
| 705 | bool AudioProcessingImpl::interleave_needed(bool is_data_processed) const { |
| 706 | // Check if we've upmixed or downmixed the audio. |
| 707 | return (num_output_channels_ != num_input_channels_ || is_data_processed); |
| 708 | } |
| 709 | |
| 710 | bool AudioProcessingImpl::synthesis_needed(bool is_data_processed) const { |
| 711 | return (is_data_processed && sample_rate_hz_ == kSampleRate32kHz); |
| 712 | } |
| 713 | |
| 714 | bool AudioProcessingImpl::analysis_needed(bool is_data_processed) const { |
| 715 | if (!is_data_processed && !voice_detection_->is_enabled()) { |
| 716 | // Only level_estimator_ is enabled. |
| 717 | return false; |
| 718 | } else if (sample_rate_hz_ == kSampleRate32kHz) { |
| 719 | // Something besides level_estimator_ is enabled, and we have super-wb. |
| 720 | return true; |
| 721 | } |
| 722 | return false; |
| 723 | } |
| 724 | |
| 725 | #ifdef WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 726 | int AudioProcessingImpl::WriteMessageToDebugFile() { |
| 727 | int32_t size = event_msg_->ByteSize(); |
| 728 | if (size <= 0) { |
| 729 | return kUnspecifiedError; |
| 730 | } |
andrew@webrtc.org | d7e9041 | 2013-10-22 10:27:23 +0000 | [diff] [blame] | 731 | #if defined(WEBRTC_ARCH_BIG_ENDIAN) |
andrew@webrtc.org | b015cbe | 2012-10-22 18:19:23 +0000 | [diff] [blame] | 732 | // TODO(ajm): Use little-endian "on the wire". For the moment, we can be |
| 733 | // pretty safe in assuming little-endian. |
| 734 | #endif |
| 735 | |
| 736 | if (!event_msg_->SerializeToString(&event_str_)) { |
| 737 | return kUnspecifiedError; |
| 738 | } |
| 739 | |
| 740 | // Write message preceded by its size. |
| 741 | if (!debug_file_->Write(&size, sizeof(int32_t))) { |
| 742 | return kFileError; |
| 743 | } |
| 744 | if (!debug_file_->Write(event_str_.data(), event_str_.length())) { |
| 745 | return kFileError; |
| 746 | } |
| 747 | |
| 748 | event_msg_->Clear(); |
| 749 | |
| 750 | return 0; |
| 751 | } |
| 752 | |
| 753 | int AudioProcessingImpl::WriteInitMessage() { |
| 754 | event_msg_->set_type(audioproc::Event::INIT); |
| 755 | audioproc::Init* msg = event_msg_->mutable_init(); |
| 756 | msg->set_sample_rate(sample_rate_hz_); |
| 757 | msg->set_device_sample_rate(echo_cancellation_->device_sample_rate_hz()); |
| 758 | msg->set_num_input_channels(num_input_channels_); |
| 759 | msg->set_num_output_channels(num_output_channels_); |
| 760 | msg->set_num_reverse_channels(num_reverse_channels_); |
| 761 | |
| 762 | int err = WriteMessageToDebugFile(); |
| 763 | if (err != kNoError) { |
| 764 | return err; |
| 765 | } |
| 766 | |
| 767 | return kNoError; |
| 768 | } |
| 769 | #endif // WEBRTC_AUDIOPROC_DEBUG_DUMP |
| 770 | } // namespace webrtc |