OLD | NEW |
---|---|
(Empty) | |
1 // Copyright 2013 The Chromium Authors. All rights reserved. | |
2 // Use of this source code is governed by a BSD-style license that can be | |
3 // found in the LICENSE file. | |
4 | |
5 #include "content/renderer/media/webrtc_audio_processor.h" | |
6 | |
7 #include "base/command_line.h" | |
8 #include "base/debug/trace_event.h" | |
9 #include "content/public/common/content_switches.h" | |
10 #include "content/renderer/media/webrtc_audio_processor_options.h" | |
11 #include "media/audio/audio_parameters.h" | |
12 #include "media/base/audio_converter.h" | |
13 #include "media/base/audio_fifo.h" | |
14 #include "media/base/channel_layout.h" | |
15 | |
16 namespace content { | |
17 | |
18 namespace { | |
19 | |
20 using webrtc::AudioProcessing; | |
21 using webrtc::MediaConstraintsInterface; | |
22 | |
23 #if defined(ANDROID) | |
24 const int kAudioProcessingSampleRate = 16000; | |
25 #else | |
26 const int kAudioProcessingSampleRate = 32000; | |
27 #endif | |
28 const int kAudioProcessingNumberOfChannel = 1; | |
29 | |
30 const int kMaxNumberOfBuffersInFifo = 2; | |
31 | |
32 } // namespace | |
33 | |
34 class WebRtcAudioProcessor::WebRtcAudioConverter | |
35 : public media::AudioConverter::InputCallback { | |
36 public: | |
37 WebRtcAudioConverter(const media::AudioParameters& source_params, | |
38 const media::AudioParameters& sink_params) | |
39 : worker_thread_detach_(false), | |
40 source_params_(source_params), | |
41 sink_params_(sink_params), | |
42 audio_converter_(source_params, sink_params_, false) { | |
43 audio_converter_.AddInput(this); | |
44 // Create and initialize audio fifo and audio bus wrapper. | |
45 // The size of the FIFO should be at least twice of the source buffer size | |
46 // or twice of the sink buffer size. | |
47 int buffer_size = std::max( | |
48 kMaxNumberOfBuffersInFifo * source_params_.frames_per_buffer(), | |
49 kMaxNumberOfBuffersInFifo * sink_params_.frames_per_buffer()); | |
50 fifo_.reset(new media::AudioFifo(source_params_.channels(), buffer_size)); | |
51 // TODO(xians): Use CreateWrapper to save one memcpy. | |
52 audio_wrapper_ = media::AudioBus::Create(sink_params_.channels(), | |
53 sink_params_.frames_per_buffer()); | |
54 } | |
55 | |
56 virtual ~WebRtcAudioConverter() { | |
57 DCHECK(destructor_thread_checker_.CalledOnValidThread()); | |
58 audio_converter_.RemoveInput(this); | |
59 } | |
60 | |
61 void Push(media::AudioBus* audio_source) { | |
62 // Called on the audio thread, which is the capture audio thread for | |
63 // |WebRtcAudioProcessor::capture_converter_|, and render audio thread for | |
64 // |WebRtcAudioProcessor::render_converter_|. | |
65 // And it must be the same thread as calling Convert(). | |
66 if (!worker_thread_detach_) { | |
67 worker_thread_checker_.DetachFromThread(); | |
Jói
2013/11/08 13:36:40
You can simply do this from the constructor I thin
no longer working on chromium
2013/11/08 15:39:30
Done, and changed the name of destructor_thread_ch
| |
68 worker_thread_detach_ = true; | |
69 } | |
70 | |
71 worker_thread_checker_.CalledOnValidThread(); | |
72 fifo_->Push(audio_source); | |
73 } | |
74 | |
75 bool Convert(webrtc::AudioFrame* out) { | |
76 // Called on the audio thread, which is the capture audio thread for | |
77 // |WebRtcAudioProcessor::capture_converter_|, and render audio thread for | |
78 // |WebRtcAudioProcessor::render_converter_|. | |
79 // Return false if there is no 10ms data in the FIFO. | |
80 worker_thread_checker_.CalledOnValidThread(); | |
81 if (fifo_->frames() < (source_params_.sample_rate() / 100)) | |
82 return false; | |
83 | |
84 // Convert 10ms data to the output format, this will trigger ProvideInput(). | |
85 audio_converter_.Convert(audio_wrapper_.get()); | |
86 | |
87 // TODO(xians): Figure out a better way to handle the interleaved and | |
88 // deinterleaved format switching. | |
89 audio_wrapper_->ToInterleaved(audio_wrapper_->frames(), | |
90 sink_params_.bits_per_sample() / 8, | |
91 out->data_); | |
92 | |
93 out->samples_per_channel_ = sink_params_.frames_per_buffer(); | |
94 out->sample_rate_hz_ = sink_params_.sample_rate(); | |
95 out->speech_type_ = webrtc::AudioFrame::kNormalSpeech; | |
96 out->vad_activity_ = webrtc::AudioFrame::kVadUnknown; | |
97 out->num_channels_ = sink_params_.channels(); | |
98 | |
99 return true; | |
100 } | |
101 | |
102 const media::AudioParameters& source_parameters() const { | |
103 return source_params_; | |
104 } | |
105 const media::AudioParameters& sink_parameters() const { | |
106 return sink_params_; | |
107 } | |
108 | |
109 private: | |
110 // AudioConverter::InputCallback implementation. | |
111 virtual double ProvideInput(media::AudioBus* audio_bus, | |
112 base::TimeDelta buffer_delay) { | |
113 // Called on realtime audio thread. | |
114 // TODO(xians): Figure out why the first Convert() triggers ProvideInput | |
115 // two times. | |
116 if (fifo_->frames() < audio_bus->frames()) | |
117 return 0; | |
118 | |
119 fifo_->Consume(audio_bus, 0, audio_bus->frames()); | |
120 return 1.0; | |
121 } | |
122 | |
123 base::ThreadChecker destructor_thread_checker_; | |
124 base::ThreadChecker worker_thread_checker_; | |
125 bool worker_thread_detach_; | |
126 media::AudioParameters source_params_; | |
127 media::AudioParameters sink_params_; | |
128 | |
129 // TODO(xians): consider using SincResampler to save some memcpy. | |
130 // Handles mixing and resampling between input and output parameters. | |
131 media::AudioConverter audio_converter_; | |
132 scoped_ptr<media::AudioBus> audio_wrapper_; | |
133 scoped_ptr<media::AudioFifo> fifo_; | |
134 }; | |
135 | |
136 WebRtcAudioProcessor::WebRtcAudioProcessor( | |
137 const webrtc::MediaConstraintsInterface* constraints) | |
138 : render_delay_ms_(0), | |
139 capture_thread_detach_(false), | |
140 render_thread_detach_(false) { | |
141 InitializeAudioProcessingModule(constraints); | |
142 } | |
143 | |
144 WebRtcAudioProcessor::~WebRtcAudioProcessor() { | |
145 DCHECK(thread_checker_.CalledOnValidThread()); | |
146 StopAudioProcessing(); | |
147 } | |
148 | |
149 void WebRtcAudioProcessor::SetCaptureFormat( | |
150 const media::AudioParameters& source_params) { | |
151 DCHECK(thread_checker_.CalledOnValidThread()); | |
152 DCHECK(source_params.IsValid()); | |
153 | |
154 // Create and initialize audio converter for the source data. | |
155 // When the webrtc AudioProcessing is enabled, the sink format of the | |
156 // converter will be the same as the post-processed data format, which is | |
157 // 32k mono for desktops and 16k mono for Android. When the AudioProcessing | |
158 // is disabled, the sink format will be the same as the source format. | |
159 const int sink_sample_rate = audio_processing_ ? | |
160 kAudioProcessingSampleRate : source_params.sample_rate(); | |
161 const media::ChannelLayout sink_channel_layout = audio_processing_ ? | |
162 media::CHANNEL_LAYOUT_MONO : source_params.channel_layout(); | |
163 | |
164 // WebRtc is using 10ms data as its native packet size. | |
165 media::AudioParameters sink_params( | |
166 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, sink_channel_layout, | |
167 sink_sample_rate, 16, sink_sample_rate / 100); | |
168 capture_converter_.reset( | |
169 new WebRtcAudioConverter(source_params, sink_params)); | |
170 } | |
171 | |
172 void WebRtcAudioProcessor::PushCaptureData(media::AudioBus* audio_source) { | |
173 if (!capture_thread_detach_) { | |
174 render_thread_checker_.DetachFromThread(); | |
Jói
2013/11/08 13:36:40
I think you meant capture_thread_checker_.DetachFr
no longer working on chromium
2013/11/08 15:39:30
Right, it should be capture_thread_checker_. Sorry
| |
175 capture_thread_detach_ = true; | |
176 } | |
177 capture_thread_checker_.CalledOnValidThread(); | |
178 capture_converter_->Push(audio_source); | |
179 } | |
180 | |
181 bool WebRtcAudioProcessor::ProcessAndConsumeData( | |
182 int capture_audio_delay_ms, int volume, bool key_pressed, | |
183 int16** out) { | |
184 capture_thread_checker_.CalledOnValidThread(); | |
185 TRACE_EVENT0("audio", | |
186 "WebRtcAudioProcessor::ProcessAndConsumeData"); | |
187 | |
188 if (!capture_converter_->Convert(&capture_frame_)) | |
189 return false; | |
190 | |
191 ProcessData(&capture_frame_, capture_audio_delay_ms, volume, key_pressed); | |
192 *out = capture_frame_.data_; | |
193 | |
194 return true; | |
195 } | |
196 | |
197 const media::AudioParameters& WebRtcAudioProcessor::OutputFormat() const { | |
198 return capture_converter_->sink_parameters(); | |
199 } | |
200 | |
201 void WebRtcAudioProcessor::ProcessData(webrtc::AudioFrame* audio_frame, | |
202 int capture_audio_delay_ms, | |
203 int volume, | |
204 bool key_pressed) { | |
205 capture_thread_checker_.CalledOnValidThread(); | |
206 if (!audio_processing_) | |
207 return; | |
208 | |
209 TRACE_EVENT0("audio", "WebRtcAudioProcessor::Process10MsData"); | |
210 DCHECK_EQ(audio_processing_->sample_rate_hz(), | |
211 capture_converter_->sink_parameters().sample_rate()); | |
212 DCHECK_EQ(audio_processing_->num_input_channels(), | |
213 capture_converter_->sink_parameters().channels()); | |
214 DCHECK_EQ(audio_processing_->num_output_channels(), | |
215 capture_converter_->sink_parameters().channels()); | |
216 | |
217 int total_delay_ms = 0; | |
218 { | |
219 base::AutoLock auto_lock(lock_); | |
220 total_delay_ms = capture_audio_delay_ms + render_delay_ms_; | |
221 } | |
222 | |
223 audio_processing_->set_stream_delay_ms(total_delay_ms); | |
224 webrtc::GainControl* agc = audio_processing_->gain_control(); | |
225 if (agc->set_stream_analog_level(volume)) | |
226 NOTREACHED(); | |
227 int err = audio_processing_->ProcessStream(audio_frame); | |
228 DCHECK(!err) << "ProcessStream() error: " << err; | |
229 | |
230 // TODO(xians): Add support for AGC, typing detectin, audio level calculation, | |
231 // stereo swapping. | |
232 } | |
233 | |
234 void WebRtcAudioProcessor::PushRenderData( | |
235 const int16* render_audio, int sample_rate, int number_of_channels, | |
236 int number_of_frames, int render_delay_ms) { | |
237 if (!render_thread_detach_) { | |
238 render_thread_checker_.DetachFromThread(); | |
Jói
2013/11/08 13:36:40
I think you can simply call this in the constructo
no longer working on chromium
2013/11/08 15:39:30
Done.
| |
239 render_thread_detach_ = true; | |
240 } | |
241 render_thread_checker_.CalledOnValidThread(); | |
242 | |
243 // Return immediately if the echo cancellation is off. | |
244 if (!audio_processing_ || | |
245 !audio_processing_->echo_cancellation()->is_enabled()) | |
246 return; | |
247 | |
248 TRACE_EVENT0("audio", | |
249 "WebRtcAudioProcessor::FeedRenderDataToAudioProcessing"); | |
250 { | |
251 base::AutoLock auto_lock(lock_); | |
252 render_delay_ms_ = render_delay_ms; | |
253 } | |
254 | |
255 InitializeRenderConverterIfNeeded(sample_rate, number_of_channels, | |
256 number_of_frames); | |
257 | |
258 // TODO(xians): Avoid this extra interleave/deinterleave. | |
259 render_data_bus_->FromInterleaved(render_audio, | |
260 render_data_bus_->frames(), | |
261 sizeof(render_audio[0])); | |
262 render_converter_->Push(render_data_bus_.get()); | |
263 while (render_converter_->Convert(&render_frame_)) { | |
264 audio_processing_->AnalyzeReverseStream(&render_frame_); | |
265 } | |
266 } | |
267 | |
268 void WebRtcAudioProcessor::InitializeAudioProcessingModule( | |
269 const webrtc::MediaConstraintsInterface* constraints) { | |
270 if (!CommandLine::ForCurrentProcess()->HasSwitch( | |
271 switches::kEnableAudioTrackProcessing)) { | |
272 return; | |
273 } | |
274 | |
275 if (!constraints) | |
276 return; | |
277 | |
278 const bool enable_aec = GetPropertyFromConstraints( | |
279 constraints, MediaConstraintsInterface::kEchoCancellation); | |
280 const bool enable_ns = GetPropertyFromConstraints( | |
281 constraints, MediaConstraintsInterface::kNoiseSuppression); | |
282 const bool enable_high_pass_filter = GetPropertyFromConstraints( | |
283 constraints, MediaConstraintsInterface::kHighpassFilter); | |
284 const bool start_aec_dump = GetPropertyFromConstraints( | |
285 constraints, MediaConstraintsInterface::kInternalAecDump); | |
286 #if defined(IOS) || defined(ANDROID) | |
287 const bool enable_experimental_aec = false; | |
288 const bool enable_typing_detection = false; | |
289 #else | |
290 const bool enable_experimental_aec = GetPropertyFromConstraints( | |
291 constraints, MediaConstraintsInterface::kExperimentalEchoCancellation); | |
292 const bool enable_typing_detection = GetPropertyFromConstraints( | |
293 constraints, MediaConstraintsInterface::kTypingNoiseDetection); | |
294 #endif | |
295 | |
296 // Reset the audio processing to NULL if no audio processing component is | |
297 // enabled. | |
298 if (!enable_aec && !enable_experimental_aec && !enable_ns && | |
299 !enable_high_pass_filter && !enable_typing_detection) { | |
300 return; | |
301 } | |
302 | |
303 // Create and configure the audio processing if it does not exist. | |
304 if (!audio_processing_) | |
305 audio_processing_.reset(webrtc::AudioProcessing::Create(0)); | |
306 | |
307 // Enable the audio processing components. | |
308 if (enable_aec) { | |
309 EnableEchoCancellation(audio_processing_.get()); | |
310 | |
311 if (enable_experimental_aec) | |
312 EnableExperimentalEchoCancellation(audio_processing_.get()); | |
313 } | |
314 | |
315 if (enable_ns) | |
316 EnableNoiseSuppression(audio_processing_.get()); | |
317 | |
318 if (enable_high_pass_filter) | |
319 EnableHighPassFilter(audio_processing_.get()); | |
320 | |
321 if (enable_typing_detection) | |
322 EnableTypingDetection(audio_processing_.get()); | |
323 | |
324 if (enable_aec && start_aec_dump) | |
325 StartAecDump(audio_processing_.get()); | |
326 | |
327 // Configure the audio format the audio processing is running on. This | |
328 // has to be done after all the needed components are enabled. | |
329 if (audio_processing_->set_sample_rate_hz(kAudioProcessingSampleRate)) | |
330 NOTREACHED(); | |
331 if (audio_processing_->set_num_channels(kAudioProcessingNumberOfChannel, | |
332 kAudioProcessingNumberOfChannel)) | |
333 NOTREACHED(); | |
334 } | |
335 | |
336 void WebRtcAudioProcessor::InitializeRenderConverterIfNeeded( | |
337 int sample_rate, int number_of_channels, int frames_per_buffer) { | |
338 // TODO, figure out if we need to handle the buffer size change. | |
Jói
2013/11/08 13:36:40
TODO needs an owner.
no longer working on chromium
2013/11/08 15:39:30
Done.
| |
339 if (render_converter_.get() && | |
340 render_converter_->source_parameters().sample_rate() == sample_rate && | |
341 render_converter_->source_parameters().channels() == number_of_channels) { | |
342 // Do nothing if the |render_converter_| has been setup properly. | |
343 return; | |
344 } | |
345 | |
346 media::AudioParameters source_params( | |
347 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, | |
348 media::GuessChannelLayout(number_of_channels), sample_rate, 16, | |
349 frames_per_buffer); | |
350 media::AudioParameters sink_params( | |
351 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, | |
352 media::CHANNEL_LAYOUT_MONO, kAudioProcessingSampleRate, 16, | |
353 kAudioProcessingSampleRate / 100); | |
354 render_converter_.reset(new WebRtcAudioConverter(source_params, sink_params)); | |
355 render_data_bus_ = media::AudioBus::Create(number_of_channels, | |
356 frames_per_buffer); | |
357 } | |
358 | |
359 void WebRtcAudioProcessor::StopAudioProcessing() { | |
360 if (!audio_processing_.get()) | |
361 return; | |
362 | |
363 // It is safe to stop the AEC dump even it is not started. | |
364 StopAecDump(audio_processing_.get()); | |
365 | |
366 audio_processing_.reset(); | |
367 } | |
368 | |
369 } // namespace content | |
OLD | NEW |