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 : source_params_(source_params), | |
40 sink_params_(sink_params), | |
41 audio_converter_(source_params, sink_params_, false) { | |
42 worker_thread_checker_.DetachFromThread(); | |
Henrik Grunell
2013/11/18 13:19:34
How does the threading model look like? Push, Conv
no longer working on chromium
2013/11/18 17:41:46
The capture converter is created in the main rende
Henrik Grunell
2013/11/19 08:52:22
Ah of course, there's two thread checkers.
| |
43 | |
44 audio_converter_.AddInput(this); | |
45 // Create and initialize audio fifo and audio bus wrapper. | |
46 // The size of the FIFO should be at least twice of the source buffer size | |
47 // or twice of the sink buffer size. | |
48 int buffer_size = std::max( | |
49 kMaxNumberOfBuffersInFifo * source_params_.frames_per_buffer(), | |
50 kMaxNumberOfBuffersInFifo * sink_params_.frames_per_buffer()); | |
51 fifo_.reset(new media::AudioFifo(source_params_.channels(), buffer_size)); | |
52 // TODO(xians): Use CreateWrapper to save one memcpy. | |
53 audio_wrapper_ = media::AudioBus::Create(sink_params_.channels(), | |
54 sink_params_.frames_per_buffer()); | |
55 } | |
56 | |
57 virtual ~WebRtcAudioConverter() { | |
58 DCHECK(create_thread_checker_.CalledOnValidThread()); | |
59 audio_converter_.RemoveInput(this); | |
60 } | |
61 | |
62 void Push(media::AudioBus* audio_source) { | |
63 // Called on the audio thread, which is the capture audio thread for | |
64 // |WebRtcAudioProcessor::capture_converter_|, and render audio thread for | |
65 // |WebRtcAudioProcessor::render_converter_|. | |
66 // And it must be the same thread as calling Convert(). | |
67 worker_thread_checker_.CalledOnValidThread(); | |
Henrik Grunell
2013/11/18 13:19:34
I may be ignorant regarding ThreadChecker, but don
no longer working on chromium
2013/11/18 17:41:46
right, it is needed.
| |
68 fifo_->Push(audio_source); | |
69 } | |
70 | |
71 bool Convert(webrtc::AudioFrame* out) { | |
72 // Called on the audio thread, which is the capture audio thread for | |
73 // |WebRtcAudioProcessor::capture_converter_|, and render audio thread for | |
74 // |WebRtcAudioProcessor::render_converter_|. | |
75 // Return false if there is no 10ms data in the FIFO. | |
76 worker_thread_checker_.CalledOnValidThread(); | |
77 if (fifo_->frames() < (source_params_.sample_rate() / 100)) | |
78 return false; | |
79 | |
80 // Convert 10ms data to the output format, this will trigger ProvideInput(). | |
81 audio_converter_.Convert(audio_wrapper_.get()); | |
82 | |
83 // TODO(xians): Figure out a better way to handle the interleaved and | |
84 // deinterleaved format switching. | |
85 audio_wrapper_->ToInterleaved(audio_wrapper_->frames(), | |
86 sink_params_.bits_per_sample() / 8, | |
87 out->data_); | |
88 | |
89 out->samples_per_channel_ = sink_params_.frames_per_buffer(); | |
90 out->sample_rate_hz_ = sink_params_.sample_rate(); | |
91 out->speech_type_ = webrtc::AudioFrame::kNormalSpeech; | |
92 out->vad_activity_ = webrtc::AudioFrame::kVadUnknown; | |
93 out->num_channels_ = sink_params_.channels(); | |
94 | |
95 return true; | |
96 } | |
97 | |
98 const media::AudioParameters& source_parameters() const { | |
99 return source_params_; | |
100 } | |
101 const media::AudioParameters& sink_parameters() const { | |
102 return sink_params_; | |
103 } | |
104 | |
105 private: | |
106 // AudioConverter::InputCallback implementation. | |
107 virtual double ProvideInput(media::AudioBus* audio_bus, | |
108 base::TimeDelta buffer_delay) OVERRIDE { | |
109 // Called on realtime audio thread. | |
110 // TODO(xians): Figure out why the first Convert() triggers ProvideInput | |
111 // two times. | |
112 if (fifo_->frames() < audio_bus->frames()) | |
113 return 0; | |
114 | |
115 fifo_->Consume(audio_bus, 0, audio_bus->frames()); | |
116 return 1.0; | |
117 } | |
118 | |
119 base::ThreadChecker create_thread_checker_; | |
120 base::ThreadChecker worker_thread_checker_; | |
121 media::AudioParameters source_params_; | |
122 media::AudioParameters sink_params_; | |
123 | |
124 // TODO(xians): consider using SincResampler to save some memcpy. | |
125 // Handles mixing and resampling between input and output parameters. | |
126 media::AudioConverter audio_converter_; | |
127 scoped_ptr<media::AudioBus> audio_wrapper_; | |
128 scoped_ptr<media::AudioFifo> fifo_; | |
129 }; | |
130 | |
131 WebRtcAudioProcessor::WebRtcAudioProcessor( | |
132 const webrtc::MediaConstraintsInterface* constraints) | |
133 : render_delay_ms_(0) { | |
134 capture_thread_checker_.DetachFromThread(); | |
135 render_thread_checker_.DetachFromThread(); | |
136 InitializeAudioProcessingModule(constraints); | |
137 } | |
138 | |
139 WebRtcAudioProcessor::~WebRtcAudioProcessor() { | |
140 DCHECK(main_thread_checker_.CalledOnValidThread()); | |
141 StopAudioProcessing(); | |
142 } | |
143 | |
144 void WebRtcAudioProcessor::SetCaptureFormat( | |
145 const media::AudioParameters& source_params) { | |
146 DCHECK(main_thread_checker_.CalledOnValidThread()); | |
147 DCHECK(source_params.IsValid()); | |
148 | |
149 // Create and initialize audio converter for the source data. | |
150 // When the webrtc AudioProcessing is enabled, the sink format of the | |
151 // converter will be the same as the post-processed data format, which is | |
152 // 32k mono for desktops and 16k mono for Android. When the AudioProcessing | |
153 // is disabled, the sink format will be the same as the source format. | |
154 const int sink_sample_rate = audio_processing_ ? | |
155 kAudioProcessingSampleRate : source_params.sample_rate(); | |
156 const media::ChannelLayout sink_channel_layout = audio_processing_ ? | |
157 media::CHANNEL_LAYOUT_MONO : source_params.channel_layout(); | |
158 | |
159 // WebRtc is using 10ms data as its native packet size. | |
160 media::AudioParameters sink_params( | |
161 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, sink_channel_layout, | |
162 sink_sample_rate, 16, sink_sample_rate / 100); | |
163 capture_converter_.reset( | |
164 new WebRtcAudioConverter(source_params, sink_params)); | |
165 } | |
166 | |
167 void WebRtcAudioProcessor::PushCaptureData(media::AudioBus* audio_source) { | |
168 capture_thread_checker_.CalledOnValidThread(); | |
Henrik Grunell
2013/11/18 13:19:34
Shouldn't you return if |audio_processor_| == NULL
no longer working on chromium
2013/11/18 17:41:46
No, we handle the case when audio_processor_ is NU
Henrik Grunell
2013/11/19 08:52:22
OK.
| |
169 capture_converter_->Push(audio_source); | |
170 } | |
171 | |
172 bool WebRtcAudioProcessor::ProcessAndConsumeData( | |
173 base::TimeDelta capture_delay, int volume, bool key_pressed, | |
174 int16** out) { | |
175 capture_thread_checker_.CalledOnValidThread(); | |
176 TRACE_EVENT0("audio", | |
177 "WebRtcAudioProcessor::ProcessAndConsumeData"); | |
178 | |
179 if (!capture_converter_->Convert(&capture_frame_)) | |
180 return false; | |
181 | |
182 ProcessData(&capture_frame_, capture_delay, volume, key_pressed); | |
183 *out = capture_frame_.data_; | |
184 | |
185 return true; | |
186 } | |
187 | |
188 const media::AudioParameters& WebRtcAudioProcessor::OutputFormat() const { | |
189 return capture_converter_->sink_parameters(); | |
190 } | |
191 | |
192 void WebRtcAudioProcessor::ProcessData(webrtc::AudioFrame* audio_frame, | |
Henrik Grunell
2013/11/18 13:19:34
Does it make sense to set the delay, volume and ke
no longer working on chromium
2013/11/18 17:41:46
They are tied to the audio_frame. From the code pe
Henrik Grunell
2013/11/19 08:52:22
OK, should they be added to AudioFrame? Or are the
no longer working on chromium
2013/11/21 15:59:25
They are only used here, we need to pass them to w
| |
193 base::TimeDelta capture_delay, | |
194 int volume, | |
195 bool key_pressed) { | |
196 capture_thread_checker_.CalledOnValidThread(); | |
197 if (!audio_processing_) | |
198 return; | |
199 | |
200 TRACE_EVENT0("audio", "WebRtcAudioProcessor::Process10MsData"); | |
201 DCHECK_EQ(audio_processing_->sample_rate_hz(), | |
202 capture_converter_->sink_parameters().sample_rate()); | |
203 DCHECK_EQ(audio_processing_->num_input_channels(), | |
204 capture_converter_->sink_parameters().channels()); | |
205 DCHECK_EQ(audio_processing_->num_output_channels(), | |
206 capture_converter_->sink_parameters().channels()); | |
207 | |
208 base::subtle::Atomic32 render_delay_ms = | |
209 base::subtle::Acquire_Load(&render_delay_ms_); | |
210 int64 capture_delay_ms = capture_delay.InMilliseconds(); | |
211 DCHECK_LT(capture_delay_ms, | |
212 std::numeric_limits<base::subtle::Atomic32>::max()); | |
213 int total_delay_ms = capture_delay_ms + render_delay_ms; | |
214 if (total_delay_ms > 1000) { | |
215 LOG(WARNING) << "Large audio delay, capture delay: " << capture_delay_ms | |
216 << "ms; render delay: " << render_delay_ms << "ms"; | |
217 } | |
218 | |
219 audio_processing_->set_stream_delay_ms(total_delay_ms); | |
220 webrtc::GainControl* agc = audio_processing_->gain_control(); | |
221 if (agc->set_stream_analog_level(volume)) | |
222 NOTREACHED(); | |
223 int err = audio_processing_->ProcessStream(audio_frame); | |
224 DCHECK(!err) << "ProcessStream() error: " << err; | |
225 | |
226 // TODO(xians): Add support for AGC, typing detection, audio level | |
227 // calculation, stereo swapping. | |
228 } | |
229 | |
230 void WebRtcAudioProcessor::PushRenderData( | |
231 const int16* render_audio, int sample_rate, int number_of_channels, | |
232 int number_of_frames, base::TimeDelta render_delay) { | |
233 render_thread_checker_.CalledOnValidThread(); | |
234 | |
235 // Return immediately if the echo cancellation is off. | |
236 if (!audio_processing_ || | |
237 !audio_processing_->echo_cancellation()->is_enabled()) | |
238 return; | |
239 | |
240 TRACE_EVENT0("audio", | |
241 "WebRtcAudioProcessor::FeedRenderDataToAudioProcessing"); | |
242 int64 new_render_delay_ms = render_delay.InMilliseconds(); | |
243 DCHECK_LT(new_render_delay_ms, | |
244 std::numeric_limits<base::subtle::Atomic32>::max()); | |
245 base::subtle::Release_Store(&render_delay_ms_, new_render_delay_ms); | |
246 | |
247 InitializeRenderConverterIfNeeded(sample_rate, number_of_channels, | |
248 number_of_frames); | |
249 | |
250 // TODO(xians): Avoid this extra interleave/deinterleave. | |
251 render_data_bus_->FromInterleaved(render_audio, | |
252 render_data_bus_->frames(), | |
253 sizeof(render_audio[0])); | |
254 render_converter_->Push(render_data_bus_.get()); | |
255 while (render_converter_->Convert(&render_frame_)) { | |
256 audio_processing_->AnalyzeReverseStream(&render_frame_); | |
257 } | |
258 } | |
259 | |
260 void WebRtcAudioProcessor::InitializeAudioProcessingModule( | |
261 const webrtc::MediaConstraintsInterface* constraints) { | |
262 if (!CommandLine::ForCurrentProcess()->HasSwitch( | |
263 switches::kEnableAudioTrackProcessing)) { | |
264 return; | |
265 } | |
266 | |
267 if (!constraints) | |
268 return; | |
269 | |
270 const bool enable_aec = GetPropertyFromConstraints( | |
271 constraints, MediaConstraintsInterface::kEchoCancellation); | |
272 const bool enable_ns = GetPropertyFromConstraints( | |
273 constraints, MediaConstraintsInterface::kNoiseSuppression); | |
274 const bool enable_high_pass_filter = GetPropertyFromConstraints( | |
275 constraints, MediaConstraintsInterface::kHighpassFilter); | |
276 const bool start_aec_dump = GetPropertyFromConstraints( | |
277 constraints, MediaConstraintsInterface::kInternalAecDump); | |
278 #if defined(IOS) || defined(ANDROID) | |
279 const bool enable_experimental_aec = false; | |
280 const bool enable_typing_detection = false; | |
281 #else | |
282 const bool enable_experimental_aec = GetPropertyFromConstraints( | |
283 constraints, MediaConstraintsInterface::kExperimentalEchoCancellation); | |
284 const bool enable_typing_detection = GetPropertyFromConstraints( | |
285 constraints, MediaConstraintsInterface::kTypingNoiseDetection); | |
286 #endif | |
287 | |
288 // Return immediately if no audio processing component is enabled. | |
289 if (!enable_aec && !enable_experimental_aec && !enable_ns && | |
290 !enable_high_pass_filter && !enable_typing_detection) { | |
291 return; | |
292 } | |
293 | |
294 // Create and configure the audio processing if it does not exist. | |
295 if (!audio_processing_) | |
296 audio_processing_.reset(webrtc::AudioProcessing::Create(0)); | |
297 | |
298 // Enable the audio processing components. | |
299 if (enable_aec) { | |
300 EnableEchoCancellation(audio_processing_.get()); | |
301 if (enable_experimental_aec) | |
302 EnableExperimentalEchoCancellation(audio_processing_.get()); | |
303 } | |
304 | |
305 if (enable_ns) | |
306 EnableNoiseSuppression(audio_processing_.get()); | |
307 | |
308 if (enable_high_pass_filter) | |
309 EnableHighPassFilter(audio_processing_.get()); | |
310 | |
311 if (enable_typing_detection) | |
312 EnableTypingDetection(audio_processing_.get()); | |
313 | |
314 if (enable_aec && start_aec_dump) | |
315 StartAecDump(audio_processing_.get()); | |
316 | |
317 // Configure the audio format the audio processing is running on. This | |
318 // has to be done after all the needed components are enabled. | |
319 if (audio_processing_->set_sample_rate_hz(kAudioProcessingSampleRate)) | |
320 NOTREACHED(); | |
321 if (audio_processing_->set_num_channels(kAudioProcessingNumberOfChannel, | |
322 kAudioProcessingNumberOfChannel)) | |
323 NOTREACHED(); | |
324 } | |
325 | |
326 void WebRtcAudioProcessor::InitializeRenderConverterIfNeeded( | |
327 int sample_rate, int number_of_channels, int frames_per_buffer) { | |
328 // TODO(xians): Figure out if we need to handle the buffer size change. | |
329 if (render_converter_.get() && | |
330 render_converter_->source_parameters().sample_rate() == sample_rate && | |
331 render_converter_->source_parameters().channels() == number_of_channels) { | |
332 // Do nothing if the |render_converter_| has been setup properly. | |
333 return; | |
334 } | |
335 | |
336 media::AudioParameters source_params( | |
337 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, | |
338 media::GuessChannelLayout(number_of_channels), sample_rate, 16, | |
339 frames_per_buffer); | |
340 media::AudioParameters sink_params( | |
341 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, | |
342 media::CHANNEL_LAYOUT_MONO, kAudioProcessingSampleRate, 16, | |
343 kAudioProcessingSampleRate / 100); | |
344 render_converter_.reset(new WebRtcAudioConverter(source_params, sink_params)); | |
345 render_data_bus_ = media::AudioBus::Create(number_of_channels, | |
346 frames_per_buffer); | |
347 } | |
348 | |
349 void WebRtcAudioProcessor::StopAudioProcessing() { | |
350 if (!audio_processing_.get()) | |
351 return; | |
352 | |
353 // It is safe to stop the AEC dump even it is not started. | |
354 StopAecDump(audio_processing_.get()); | |
355 | |
356 audio_processing_.reset(); | |
357 } | |
358 | |
359 } // namespace content | |
OLD | NEW |