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_util.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 audio_converter_.AddInput(this); | |
43 // Create and initialize audio fifo and audio bus wrapper. | |
44 // The size of the FIFO should be at least twice of the source buffer size | |
45 // or twice of the sink buffer size. | |
46 int buffer_size = std::max( | |
47 kMaxNumberOfBuffersInFifo * source_params_.frames_per_buffer(), | |
48 kMaxNumberOfBuffersInFifo * sink_params_.frames_per_buffer()); | |
49 fifo_.reset(new media::AudioFifo(source_params_.channels(), buffer_size)); | |
50 // TODO(xians): Use CreateWrapper to save one memcpy. | |
51 audio_wrapper_ = media::AudioBus::Create(sink_params_.channels(), | |
52 sink_params_.frames_per_buffer()); | |
53 } | |
54 | |
55 virtual ~WebRtcAudioConverter() { | |
56 DCHECK(thread_checker_.CalledOnValidThread()); | |
57 audio_converter_.RemoveInput(this); | |
58 } | |
59 | |
60 void Push(media::AudioBus* audio_source) { | |
61 // Called on the realtime audio thread, which must be the same thread as | |
62 // calling Convert(). | |
DaleCurtis
2013/11/07 20:44:08
Isn't this called on the capture thread while Conv
no longer working on chromium
2013/11/08 13:01:15
There are two Converters used by the processor: ca
| |
63 fifo_->Push(audio_source); | |
64 } | |
65 | |
66 bool Convert(webrtc::AudioFrame* out) { | |
67 // Called on realtime audio thread. | |
68 // Return false if there is no 10ms data in the FIFO. | |
69 if (fifo_->frames() < (source_params_.sample_rate() / 100)) | |
70 return false; | |
71 | |
72 // Convert 10ms data to the output format, this will trigger ProvideInput(). | |
73 audio_converter_.Convert(audio_wrapper_.get()); | |
74 | |
75 // TODO(xians): Figure out a better way to handle the interleaved and | |
76 // deinterleaved format switching. | |
77 audio_wrapper_->ToInterleaved(audio_wrapper_->frames(), | |
78 sink_params_.bits_per_sample() / 8, | |
79 out->data_); | |
80 | |
81 out->samples_per_channel_ = sink_params_.frames_per_buffer(); | |
82 out->sample_rate_hz_ = sink_params_.sample_rate(); | |
83 out->speech_type_ = webrtc::AudioFrame::kNormalSpeech; | |
84 out->vad_activity_ = webrtc::AudioFrame::kVadUnknown; | |
85 out->num_channels_ = sink_params_.channels(); | |
86 | |
87 return true; | |
88 } | |
89 | |
90 const media::AudioParameters& source_parameters() const { | |
91 return source_params_; | |
92 } | |
93 const media::AudioParameters& sink_parameters() const { | |
94 return sink_params_; | |
95 } | |
96 | |
97 private: | |
98 // AudioConverter::InputCallback implementation. | |
99 virtual double ProvideInput(media::AudioBus* audio_bus, | |
100 base::TimeDelta buffer_delay) { | |
101 // Called on realtime audio thread. | |
DaleCurtis
2013/11/07 20:44:08
Ditto.
no longer working on chromium
2013/11/08 13:01:15
The same.
| |
102 // TODO(xians): Figure out why the first Convert() triggers ProvideInput | |
103 // two times. | |
104 if (fifo_->frames() < audio_bus->frames()) | |
105 return 0; | |
106 | |
107 fifo_->Consume(audio_bus, 0, audio_bus->frames()); | |
108 return 1.0; | |
109 } | |
110 | |
111 base::ThreadChecker thread_checker_; | |
112 media::AudioParameters source_params_; | |
113 media::AudioParameters sink_params_; | |
114 | |
115 // TODO(xians): consider using SincResampler to save some memcpy. | |
116 // Handles mixing and resampling between input and output parameters. | |
117 media::AudioConverter audio_converter_; | |
118 scoped_ptr<media::AudioBus> audio_wrapper_; | |
119 scoped_ptr<media::AudioFifo> fifo_; | |
120 }; | |
121 | |
122 WebRtcAudioProcessor::WebRtcAudioProcessor( | |
123 const webrtc::MediaConstraintsInterface* constraints) | |
124 : render_delay_ms_(0) { | |
125 InitializeAudioProcessingModule(constraints); | |
126 } | |
127 | |
128 WebRtcAudioProcessor::~WebRtcAudioProcessor() { | |
129 DCHECK(thread_checker_.CalledOnValidThread()); | |
130 StopAudioProcessing(); | |
131 } | |
132 | |
133 void WebRtcAudioProcessor::SetCaptureFormat( | |
134 const media::AudioParameters& source_params) { | |
135 DCHECK(thread_checker_.CalledOnValidThread()); | |
136 DCHECK(source_params.IsValid()); | |
137 | |
138 // Create and initialize audio converter for the source data. | |
139 // When the webrtc AudioProcessing is enabled, the sink format of the | |
140 // converter will be the same as the post-processed data format, which is | |
141 // 32k mono for desktops and 16k mono for Android. When the AudioProcessing | |
142 // is disabled, the sink format will be the same as the source format. | |
143 const int sink_sample_rate = audio_processing_ ? | |
144 kAudioProcessingSampleRate : source_params.sample_rate(); | |
145 const media::ChannelLayout sink_channel_layout = audio_processing_ ? | |
146 media::CHANNEL_LAYOUT_MONO : source_params.channel_layout(); | |
147 | |
148 // WebRtc is using 10ms data as its native packet size. | |
149 media::AudioParameters sink_params( | |
150 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, sink_channel_layout, | |
151 sink_sample_rate, 16, sink_sample_rate / 100); | |
152 capture_converter_.reset( | |
153 new WebRtcAudioConverter(source_params, sink_params)); | |
154 } | |
155 | |
156 void WebRtcAudioProcessor::PushCaptureData(media::AudioBus* audio_source) { | |
157 capture_converter_->Push(audio_source); | |
158 } | |
159 | |
160 bool WebRtcAudioProcessor::ProcessAndConsumeData( | |
161 int capture_audio_delay_ms, int volume, bool key_pressed, | |
162 int16** out) { | |
DaleCurtis
2013/11/07 20:44:08
Instead of int16**, should this just be a WebRtcAu
no longer working on chromium
2013/11/08 13:01:15
out is supposed to be used by all clients of the t
| |
163 TRACE_EVENT0("audio", | |
164 "WebRtcAudioProcessor::ProcessAndConsumeData"); | |
165 | |
166 if (!capture_converter_->Convert(&capture_frame_)) | |
167 return false; | |
168 | |
169 ProcessData(&capture_frame_, capture_audio_delay_ms, volume, key_pressed); | |
170 *out = capture_frame_.data_; | |
171 | |
172 return true; | |
173 } | |
174 | |
175 const media::AudioParameters& WebRtcAudioProcessor::OutputFormat() const { | |
176 return capture_converter_->sink_parameters(); | |
177 } | |
178 | |
179 void WebRtcAudioProcessor::ProcessData(webrtc::AudioFrame* audio_frame, | |
DaleCurtis
2013/11/07 20:44:08
As with above, maybe having a DCHECK(capture_threa
no longer working on chromium
2013/11/08 13:01:15
Done.
| |
180 int capture_audio_delay_ms, | |
181 int volume, | |
182 bool key_pressed) { | |
183 if (!audio_processing_) | |
184 return; | |
185 | |
186 TRACE_EVENT0("audio", "WebRtcAudioProcessor::Process10MsData"); | |
187 DCHECK_EQ(audio_processing_->sample_rate_hz(), | |
188 capture_converter_->sink_parameters().sample_rate()); | |
189 DCHECK_EQ(audio_processing_->num_input_channels(), | |
190 capture_converter_->sink_parameters().channels()); | |
191 DCHECK_EQ(audio_processing_->num_output_channels(), | |
192 capture_converter_->sink_parameters().channels()); | |
193 | |
194 int total_delay_ms = 0; | |
195 { | |
196 base::AutoLock auto_lock(lock_); | |
197 total_delay_ms = capture_audio_delay_ms + render_delay_ms_; | |
198 } | |
199 | |
200 audio_processing_->set_stream_delay_ms(total_delay_ms); | |
201 webrtc::GainControl* agc = audio_processing_->gain_control(); | |
202 if (agc->set_stream_analog_level(volume)) | |
203 NOTREACHED(); | |
204 int err = audio_processing_->ProcessStream(audio_frame); | |
205 DCHECK(!err) << "ProcessStream() error: " << err; | |
206 | |
207 // TODO(xians): Add support for AGC, typing detectin, audio level calculation, | |
208 // stereo swapping. | |
209 } | |
210 | |
211 void WebRtcAudioProcessor::PushRenderData( | |
212 const int16* render_audio, int sample_rate, int number_of_channels, | |
213 int number_of_frames, int render_delay_ms) { | |
214 // Return immediately if the echo cancellation is off. | |
215 if (!audio_processing_ || | |
216 !audio_processing_->echo_cancellation()->is_enabled()) | |
217 return; | |
218 | |
219 TRACE_EVENT0("audio", | |
220 "WebRtcAudioProcessor::FeedRenderDataToAudioProcessing"); | |
221 { | |
222 base::AutoLock auto_lock(lock_); | |
223 render_delay_ms_ = render_delay_ms; | |
224 } | |
225 | |
226 InitializeRenderConverterIfNeeded(sample_rate, number_of_channels, | |
227 number_of_frames); | |
228 | |
229 // TODO(xians): Avoid this extra interleave/deinterleave. | |
230 render_data_bus_->FromInterleaved(render_audio, | |
231 render_data_bus_->frames(), | |
232 sizeof(render_audio[0])); | |
233 render_converter_->Push(render_data_bus_.get()); | |
234 while (render_converter_->Convert(&render_frame_)) { | |
235 audio_processing_->AnalyzeReverseStream(&render_frame_); | |
236 } | |
237 } | |
238 | |
239 void WebRtcAudioProcessor::InitializeAudioProcessingModule( | |
240 const webrtc::MediaConstraintsInterface* constraints) { | |
241 if (!CommandLine::ForCurrentProcess()->HasSwitch( | |
242 switches::kEnableAudioTrackProcessing)) { | |
243 return; | |
244 } | |
245 | |
246 if (!constraints) | |
247 return; | |
248 | |
249 const bool enable_aec = GetPropertyFromConstraints( | |
250 constraints, MediaConstraintsInterface::kEchoCancellation); | |
251 const bool enable_ns = GetPropertyFromConstraints( | |
252 constraints, MediaConstraintsInterface::kNoiseSuppression); | |
253 const bool enable_high_pass_filter = GetPropertyFromConstraints( | |
254 constraints, MediaConstraintsInterface::kHighpassFilter); | |
255 const bool start_aec_dump = GetPropertyFromConstraints( | |
256 constraints, MediaConstraintsInterface::kInternalAecDump); | |
257 #if defined(IOS) || defined(ANDROID) | |
258 const bool enable_experimental_aec = false; | |
259 const bool enable_typing_detection = false; | |
260 #else | |
261 const bool enable_experimental_aec = GetPropertyFromConstraints( | |
262 constraints, MediaConstraintsInterface::kExperimentalEchoCancellation); | |
263 const bool enable_typing_detection = GetPropertyFromConstraints( | |
264 constraints, MediaConstraintsInterface::kTypingNoiseDetection); | |
265 #endif | |
266 | |
267 // Reset the audio processing to NULL if no audio processing component is | |
268 // enabled. | |
269 if (!enable_aec && !enable_experimental_aec && !enable_ns && | |
270 !enable_high_pass_filter && !enable_typing_detection) { | |
271 return; | |
272 } | |
273 | |
274 // Create and configure the audio processing if it does not exist. | |
275 if (!audio_processing_) | |
276 audio_processing_.reset(webrtc::AudioProcessing::Create(0)); | |
277 | |
278 // Enable the audio processing components. | |
279 if (enable_aec) { | |
280 EnableEchoCancellation(audio_processing_.get()); | |
281 | |
282 if (enable_experimental_aec) | |
283 EnableExperimentalEchoCancellation(audio_processing_.get()); | |
284 } | |
285 | |
286 if (enable_ns) | |
287 EnableNoiseSuppression(audio_processing_.get()); | |
288 | |
289 if (enable_high_pass_filter) | |
290 EnableHighPassFilter(audio_processing_.get()); | |
291 | |
292 if (enable_typing_detection) | |
293 EnableTypingDetection(audio_processing_.get()); | |
294 | |
295 if (enable_aec && start_aec_dump) | |
296 StartAecDump(audio_processing_.get()); | |
297 | |
298 // Configure the audio format the audio processing is running on. This | |
299 // has to be done after all the needed components are enabled. | |
300 if (audio_processing_->set_sample_rate_hz(kAudioProcessingSampleRate)) | |
301 NOTREACHED(); | |
302 if (audio_processing_->set_num_channels(kAudioProcessingNumberOfChannel, | |
303 kAudioProcessingNumberOfChannel)) | |
304 NOTREACHED(); | |
305 } | |
306 | |
307 void WebRtcAudioProcessor::InitializeRenderConverterIfNeeded( | |
308 int sample_rate, int number_of_channels, int frames_per_buffer) { | |
309 // TODO, figure out if we need to handle the buffer size change. | |
310 if (render_converter_.get() && | |
311 render_converter_->source_parameters().sample_rate() == sample_rate && | |
312 render_converter_->source_parameters().channels() == number_of_channels) { | |
313 // Do nothing if the |render_converter_| has been setup properly. | |
314 return; | |
315 } | |
316 | |
317 media::AudioParameters source_params( | |
318 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, | |
319 media::GuessChannelLayout(number_of_channels), sample_rate, 16, | |
320 frames_per_buffer); | |
321 media::AudioParameters sink_params( | |
322 media::AudioParameters::AUDIO_PCM_LOW_LATENCY, | |
323 media::CHANNEL_LAYOUT_MONO, kAudioProcessingSampleRate, 16, | |
324 kAudioProcessingSampleRate / 100); | |
325 render_converter_.reset(new WebRtcAudioConverter(source_params, sink_params)); | |
326 render_data_bus_ = media::AudioBus::Create(number_of_channels, | |
327 frames_per_buffer); | |
328 } | |
329 | |
330 void WebRtcAudioProcessor::StopAudioProcessing() { | |
331 if (!audio_processing_.get()) | |
332 return; | |
333 | |
334 // It is safe to stop the AEC dump even it is not started. | |
335 StopAecDump(audio_processing_.get()); | |
336 | |
337 audio_processing_.reset(); | |
338 } | |
339 | |
340 } // namespace content | |
OLD | NEW |