content/renderer/media/media_stream_audio_processor_options.cc

   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/media_stream_audio_processor_options.h"
   6
   7 #include "base/files/file_path.h"
   8 #include "base/files/file_util.h"
   9 #include "base/logging.h"
  10 #include "base/metrics/field_trial.h"
  11 #include "base/metrics/histogram.h"
  12 #include "base/strings/string_number_conversions.h"
  13 #include "base/strings/utf_string_conversions.h"
  14 #include "content/common/media/media_stream_options.h"
  15 #include "content/renderer/media/media_stream_constraints_util.h"
  16 #include "content/renderer/media/media_stream_source.h"
  17 #include "content/renderer/media/rtc_media_constraints.h"
  18 #include "media/audio/audio_parameters.h"
  19 #include "third_party/webrtc/modules/audio_processing/include/audio_processing.h"
  20 #include "third_party/webrtc/modules/audio_processing/typing_detection.h"
  21
  22 namespace content {
  23
  24 const char MediaAudioConstraints::kEchoCancellation[] = "echoCancellation";
  25 const char MediaAudioConstraints::kGoogEchoCancellation[] =
  26     "googEchoCancellation";
  27 const char MediaAudioConstraints::kGoogExperimentalEchoCancellation[] =
  28     "googEchoCancellation2";
  29 const char MediaAudioConstraints::kGoogAutoGainControl[] =
  30     "googAutoGainControl";
  31 const char MediaAudioConstraints::kGoogExperimentalAutoGainControl[] =
  32     "googAutoGainControl2";
  33 const char MediaAudioConstraints::kGoogNoiseSuppression[] =
  34     "googNoiseSuppression";
  35 const char MediaAudioConstraints::kGoogExperimentalNoiseSuppression[] =
  36     "googNoiseSuppression2";
  37 const char MediaAudioConstraints::kGoogBeamforming[] = "googBeamforming";
  38 const char MediaAudioConstraints::kGoogHighpassFilter[] = "googHighpassFilter";
  39 const char MediaAudioConstraints::kGoogTypingNoiseDetection[] =
  40     "googTypingNoiseDetection";
  41 const char MediaAudioConstraints::kGoogAudioMirroring[] = "googAudioMirroring";
  42 const char MediaAudioConstraints::kGoogAudioProcessing48kHzSupport[] =
  43     "googAudioProcessing48kHzSupport";
  44
  45 namespace {
  46
  47 // Constant constraint keys which enables default audio constraints on
  48 // mediastreams with audio.
  49 struct {
  50   const char* key;
  51   bool value;
  52 } const kDefaultAudioConstraints[] = {
  53   { MediaAudioConstraints::kEchoCancellation, true },
  54   { MediaAudioConstraints::kGoogEchoCancellation, true },
  55 #if defined(OS_ANDROID) || defined(OS_IOS)
  56   { MediaAudioConstraints::kGoogExperimentalEchoCancellation, false },
  57 #else
  58   // Enable the extended filter mode AEC on all non-mobile platforms.
  59   { MediaAudioConstraints::kGoogExperimentalEchoCancellation, true },
  60 #endif
  61   { MediaAudioConstraints::kGoogAutoGainControl, true },
  62   { MediaAudioConstraints::kGoogExperimentalAutoGainControl, true },
  63   { MediaAudioConstraints::kGoogNoiseSuppression, true },
  64   { MediaAudioConstraints::kGoogHighpassFilter, true },
  65   { MediaAudioConstraints::kGoogTypingNoiseDetection, true },
  66   { MediaAudioConstraints::kGoogExperimentalNoiseSuppression, false },
  67   { MediaAudioConstraints::kGoogBeamforming, false },
  68 #if defined(OS_WIN)
  69   { kMediaStreamAudioDucking, true },
  70 #else
  71   { kMediaStreamAudioDucking, false },
  72 #endif
  73   { kMediaStreamAudioHotword, false },
  74   { MediaAudioConstraints::kGoogAudioProcessing48kHzSupport, false },
  75 };
  76
  77 bool IsAudioProcessingConstraint(const std::string& key) {
  78   // |kMediaStreamAudioDucking| does not require audio processing.
  79   return key != kMediaStreamAudioDucking;
  80 }
  81
  82 // Used to log echo quality based on delay estimates.
  83 enum DelayBasedEchoQuality {
  84   DELAY_BASED_ECHO_QUALITY_GOOD = 0,
  85   DELAY_BASED_ECHO_QUALITY_SPURIOUS,
  86   DELAY_BASED_ECHO_QUALITY_BAD,
  87   DELAY_BASED_ECHO_QUALITY_INVALID,
  88   DELAY_BASED_ECHO_QUALITY_MAX
  89 };
  90
  91 DelayBasedEchoQuality EchoDelayFrequencyToQuality(float delay_frequency) {
  92   const float kEchoDelayFrequencyLowerLimit = 0.1f;
  93   const float kEchoDelayFrequencyUpperLimit = 0.8f;
  94   // DELAY_BASED_ECHO_QUALITY_GOOD
  95   //   delay is out of bounds during at most 10 % of the time.
  96   // DELAY_BASED_ECHO_QUALITY_SPURIOUS
  97   //   delay is out of bounds 10-80 % of the time.
  98   // DELAY_BASED_ECHO_QUALITY_BAD
  99   //   delay is mostly out of bounds >= 80 % of the time.
 100   // DELAY_BASED_ECHO_QUALITY_INVALID
 101   //   delay_frequency is negative which happens if we have insufficient data.
 102   if (delay_frequency < 0)
 103     return DELAY_BASED_ECHO_QUALITY_INVALID;
 104   else if (delay_frequency <= kEchoDelayFrequencyLowerLimit)
 105     return DELAY_BASED_ECHO_QUALITY_GOOD;
 106   else if (delay_frequency < kEchoDelayFrequencyUpperLimit)
 107     return DELAY_BASED_ECHO_QUALITY_SPURIOUS;
 108   else
 109     return DELAY_BASED_ECHO_QUALITY_BAD;
 110 }
 111
 112 }  // namespace
 113
 114 // TODO(xians): Remove this method after the APM in WebRtc is deprecated.
 115 void MediaAudioConstraints::ApplyFixedAudioConstraints(
 116     RTCMediaConstraints* constraints) {
 117   for (size_t i = 0; i < arraysize(kDefaultAudioConstraints); ++i) {
 118     bool already_set_value;
 119     if (!webrtc::FindConstraint(constraints, kDefaultAudioConstraints[i].key,
 120                                 &already_set_value, NULL)) {
 121       const std::string value = kDefaultAudioConstraints[i].value ?
 122           webrtc::MediaConstraintsInterface::kValueTrue :
 123           webrtc::MediaConstraintsInterface::kValueFalse;
 124       constraints->AddOptional(kDefaultAudioConstraints[i].key, value, false);
 125     } else {
 126       DVLOG(1) << "Constraint " << kDefaultAudioConstraints[i].key
 127                << " already set to " << already_set_value;
 128     }
 129   }
 130 }
 131
 132 MediaAudioConstraints::MediaAudioConstraints(
 133     const blink::WebMediaConstraints& constraints, int effects)
 134     : constraints_(constraints),
 135       effects_(effects),
 136       default_audio_processing_constraint_value_(true) {
 137   // The default audio processing constraints are turned off when
 138   // - gUM has a specific kMediaStreamSource, which is used by tab capture
 139   //   and screen capture.
 140   // - |kEchoCancellation| is explicitly set to false.
 141   std::string value_str;
 142   bool value_bool = false;
 143   if ((GetConstraintValueAsString(constraints, kMediaStreamSource,
 144                                   &value_str)) ||
 145       (GetConstraintValueAsBoolean(constraints_, kEchoCancellation,
 146                                    &value_bool) && !value_bool)) {
 147     default_audio_processing_constraint_value_ = false;
 148   }
 149 }
 150
 151 MediaAudioConstraints::~MediaAudioConstraints() {}
 152
 153 bool MediaAudioConstraints::GetProperty(const std::string& key) const {
 154   // Return the value if the constraint is specified in |constraints|,
 155   // otherwise return the default value.
 156   bool value = false;
 157   if (!GetConstraintValueAsBoolean(constraints_, key, &value))
 158     value = GetDefaultValueForConstraint(constraints_, key);
 159
 160   return value;
 161 }
 162
 163 bool MediaAudioConstraints::GetEchoCancellationProperty() const {
 164   // If platform echo canceller is enabled, disable the software AEC.
 165   if (effects_ & media::AudioParameters::ECHO_CANCELLER)
 166     return false;
 167
 168   // If |kEchoCancellation| is specified in the constraints, it will
 169   // override the value of |kGoogEchoCancellation|.
 170   bool value = false;
 171   if (GetConstraintValueAsBoolean(constraints_, kEchoCancellation, &value))
 172     return value;
 173
 174   return GetProperty(kGoogEchoCancellation);
 175 }
 176
 177 bool MediaAudioConstraints::IsValid() const {
 178   blink::WebVector<blink::WebMediaConstraint> mandatory;
 179   constraints_.getMandatoryConstraints(mandatory);
 180   for (size_t i = 0; i < mandatory.size(); ++i) {
 181     const std::string key = mandatory[i].m_name.utf8();
 182     if (key == kMediaStreamSource || key == kMediaStreamSourceId ||
 183         key == MediaStreamSource::kSourceId) {
 184       // Ignore Chrome specific Tab capture and |kSourceId| constraints.
 185       continue;
 186     }
 187
 188     bool valid = false;
 189     for (size_t j = 0; j < arraysize(kDefaultAudioConstraints); ++j) {
 190       if (key == kDefaultAudioConstraints[j].key) {
 191         bool value = false;
 192         valid = GetMandatoryConstraintValueAsBoolean(constraints_, key, &value);
 193         break;
 194       }
 195     }
 196
 197     if (!valid) {
 198       DLOG(ERROR) << "Invalid MediaStream constraint. Name: " << key;
 199       return false;
 200     }
 201   }
 202
 203   return true;
 204 }
 205
 206 bool MediaAudioConstraints::GetDefaultValueForConstraint(
 207     const blink::WebMediaConstraints& constraints,
 208     const std::string& key) const {
 209   // |kMediaStreamAudioDucking| is not restricted by
 210   // |default_audio_processing_constraint_value_| since it does not require
 211   // audio processing.
 212   if (!default_audio_processing_constraint_value_ &&
 213       IsAudioProcessingConstraint(key))
 214     return false;
 215
 216   for (size_t i = 0; i < arraysize(kDefaultAudioConstraints); ++i) {
 217     if (kDefaultAudioConstraints[i].key == key)
 218       return kDefaultAudioConstraints[i].value;
 219   }
 220
 221   return false;
 222 }
 223
 224 EchoInformation::EchoInformation()
 225     : num_chunks_(0) {}
 226
 227 EchoInformation::~EchoInformation() {}
 228
 229 void EchoInformation::UpdateAecDelayStats(
 230     webrtc::EchoCancellation* echo_cancellation) {
 231   // In WebRTC, three echo delay metrics are calculated and updated every
 232   // five seconds. We use one of them, |fraction_poor_delays| to log in a UMA
 233   // histogram an Echo Cancellation quality metric. The stat in WebRTC has a
 234   // fixed aggregation window of five seconds, so we use the same query
 235   // frequency to avoid logging old values.
 236   const int kNumChunksInFiveSeconds = 500;
 237   if (!echo_cancellation->is_delay_logging_enabled() ||
 238       !echo_cancellation->is_enabled()) {
 239     return;
 240   }
 241
 242   num_chunks_++;
 243   if (num_chunks_ < kNumChunksInFiveSeconds) {
 244     return;
 245   }
 246
 247   int dummy_median = 0, dummy_std = 0;
 248   float fraction_poor_delays = 0;
 249   if (echo_cancellation->GetDelayMetrics(
 250           &dummy_median, &dummy_std, &fraction_poor_delays) ==
 251       webrtc::AudioProcessing::kNoError) {
 252     num_chunks_ = 0;
 253     // Map |fraction_poor_delays| to an Echo Cancellation quality and log in UMA
 254     // histogram. See DelayBasedEchoQuality for information on histogram
 255     // buckets.
 256     UMA_HISTOGRAM_ENUMERATION("WebRTC.AecDelayBasedQuality",
 257                               EchoDelayFrequencyToQuality(fraction_poor_delays),
 258                               DELAY_BASED_ECHO_QUALITY_MAX);
 259   }
 260 }
 261
 262 void EnableEchoCancellation(AudioProcessing* audio_processing) {
 263 #if defined(OS_ANDROID) || defined(OS_IOS)
 264   const std::string group_name =
 265       base::FieldTrialList::FindFullName("ReplaceAECMWithAEC");
 266   if (group_name.empty() ||
 267       !(group_name == "Enabled" || group_name == "DefaultEnabled")) {
 268     // Mobile devices are using AECM.
 269     int err = audio_processing->echo_control_mobile()->set_routing_mode(
 270         webrtc::EchoControlMobile::kSpeakerphone);
 271     err |= audio_processing->echo_control_mobile()->Enable(true);
 272     CHECK_EQ(err, 0);
 273     return;
 274   }
 275 #endif
 276   int err = audio_processing->echo_cancellation()->set_suppression_level(
 277       webrtc::EchoCancellation::kHighSuppression);
 278
 279   // Enable the metrics for AEC.
 280   err |= audio_processing->echo_cancellation()->enable_metrics(true);
 281   err |= audio_processing->echo_cancellation()->enable_delay_logging(true);
 282   err |= audio_processing->echo_cancellation()->Enable(true);
 283   CHECK_EQ(err, 0);
 284 }
 285
 286 void EnableNoiseSuppression(AudioProcessing* audio_processing) {
 287   int err = audio_processing->noise_suppression()->set_level(
 288       webrtc::NoiseSuppression::kHigh);
 289   err |= audio_processing->noise_suppression()->Enable(true);
 290   CHECK_EQ(err, 0);
 291 }
 292
 293 void EnableHighPassFilter(AudioProcessing* audio_processing) {
 294   CHECK_EQ(audio_processing->high_pass_filter()->Enable(true), 0);
 295 }
 296
 297 void EnableTypingDetection(AudioProcessing* audio_processing,
 298                            webrtc::TypingDetection* typing_detector) {
 299   int err = audio_processing->voice_detection()->Enable(true);
 300   err |= audio_processing->voice_detection()->set_likelihood(
 301       webrtc::VoiceDetection::kVeryLowLikelihood);
 302   CHECK_EQ(err, 0);
 303
 304   // Configure the update period to 1s (100 * 10ms) in the typing detector.
 305   typing_detector->SetParameters(0, 0, 0, 0, 0, 100);
 306 }
 307
 308 void StartEchoCancellationDump(AudioProcessing* audio_processing,
 309                                base::File aec_dump_file) {
 310   DCHECK(aec_dump_file.IsValid());
 311
 312   FILE* stream = base::FileToFILE(aec_dump_file.Pass(), "w");
 313   if (!stream) {
 314     LOG(ERROR) << "Failed to open AEC dump file";
 315     return;
 316   }
 317
 318   if (audio_processing->StartDebugRecording(stream))
 319     DLOG(ERROR) << "Fail to start AEC debug recording";
 320 }
 321
 322 void StopEchoCancellationDump(AudioProcessing* audio_processing) {
 323   if (audio_processing->StopDebugRecording())
 324     DLOG(ERROR) << "Fail to stop AEC debug recording";
 325 }
 326
 327 void EnableAutomaticGainControl(AudioProcessing* audio_processing) {
 328 #if defined(OS_ANDROID) || defined(OS_IOS)
 329   const webrtc::GainControl::Mode mode = webrtc::GainControl::kFixedDigital;
 330 #else
 331   const webrtc::GainControl::Mode mode = webrtc::GainControl::kAdaptiveAnalog;
 332 #endif
 333   int err = audio_processing->gain_control()->set_mode(mode);
 334   err |= audio_processing->gain_control()->Enable(true);
 335   CHECK_EQ(err, 0);
 336 }
 337
 338 void GetAecStats(webrtc::EchoCancellation* echo_cancellation,
 339                  webrtc::AudioProcessorInterface::AudioProcessorStats* stats) {
 340   // These values can take on valid negative values, so use the lowest possible
 341   // level as default rather than -1.
 342   stats->echo_return_loss = -100;
 343   stats->echo_return_loss_enhancement = -100;
 344
 345   // The median value can also be negative, but in practice -1 is only used to
 346   // signal insufficient data, since the resolution is limited to multiples
 347   // of 4ms.
 348   stats->echo_delay_median_ms = -1;
 349   stats->echo_delay_std_ms = -1;
 350
 351   // TODO(ajm): Re-enable this metric once we have a reliable implementation.
 352   stats->aec_quality_min = -1.0f;
 353
 354   if (!echo_cancellation->are_metrics_enabled() ||
 355       !echo_cancellation->is_delay_logging_enabled() ||
 356       !echo_cancellation->is_enabled()) {
 357     return;
 358   }
 359
 360   // TODO(ajm): we may want to use VoECallReport::GetEchoMetricsSummary
 361   // here, but it appears to be unsuitable currently. Revisit after this is
 362   // investigated: http://b/issue?id=5666755
 363   webrtc::EchoCancellation::Metrics echo_metrics;
 364   if (!echo_cancellation->GetMetrics(&echo_metrics)) {
 365     stats->echo_return_loss = echo_metrics.echo_return_loss.instant;
 366     stats->echo_return_loss_enhancement =
 367         echo_metrics.echo_return_loss_enhancement.instant;
 368   }
 369
 370   int median = 0, std = 0;
 371   float dummy = 0;
 372   if (echo_cancellation->GetDelayMetrics(&median, &std, &dummy) ==
 373       webrtc::AudioProcessing::kNoError) {
 374     stats->echo_delay_median_ms = median;
 375     stats->echo_delay_std_ms = std;
 376   }
 377 }
 378
 379 }  // namespace content