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/string_split.h"
  14 #include "base/strings/string_util.h"
  15 #include "base/strings/utf_string_conversions.h"
  16 #include "content/common/media/media_stream_options.h"
  17 #include "content/renderer/media/media_stream_constraints_util.h"
  18 #include "content/renderer/media/media_stream_source.h"
  19 #include "content/renderer/media/rtc_media_constraints.h"
  20 #include "media/audio/audio_parameters.h"
  21 #include "third_party/webrtc/modules/audio_processing/include/audio_processing.h"
  22 #include "third_party/webrtc/modules/audio_processing/typing_detection.h"
  23
  24 namespace content {
  25
  26 const char MediaAudioConstraints::kEchoCancellation[] = "echoCancellation";
  27 const char MediaAudioConstraints::kGoogEchoCancellation[] =
  28     "googEchoCancellation";
  29 const char MediaAudioConstraints::kGoogExperimentalEchoCancellation[] =
  30     "googEchoCancellation2";
  31 const char MediaAudioConstraints::kGoogAutoGainControl[] =
  32     "googAutoGainControl";
  33 const char MediaAudioConstraints::kGoogExperimentalAutoGainControl[] =
  34     "googAutoGainControl2";
  35 const char MediaAudioConstraints::kGoogNoiseSuppression[] =
  36     "googNoiseSuppression";
  37 const char MediaAudioConstraints::kGoogExperimentalNoiseSuppression[] =
  38     "googNoiseSuppression2";
  39 const char MediaAudioConstraints::kGoogBeamforming[] = "googBeamforming";
  40 const char MediaAudioConstraints::kGoogArrayGeometry[] = "googArrayGeometry";
  41 const char MediaAudioConstraints::kGoogHighpassFilter[] = "googHighpassFilter";
  42 const char MediaAudioConstraints::kGoogTypingNoiseDetection[] =
  43     "googTypingNoiseDetection";
  44 const char MediaAudioConstraints::kGoogAudioMirroring[] = "googAudioMirroring";
  45
  46 namespace {
  47
  48 // Constant constraint keys which enables default audio constraints on
  49 // mediastreams with audio.
  50 struct {
  51   const char* key;
  52   bool value;
  53 } const kDefaultAudioConstraints[] = {
  54   { MediaAudioConstraints::kEchoCancellation, true },
  55   { MediaAudioConstraints::kGoogEchoCancellation, true },
  56 #if defined(OS_ANDROID) || defined(OS_IOS)
  57   { MediaAudioConstraints::kGoogExperimentalEchoCancellation, false },
  58 #else
  59   // Enable the extended filter mode AEC on all non-mobile platforms.
  60   { MediaAudioConstraints::kGoogExperimentalEchoCancellation, true },
  61 #endif
  62   { MediaAudioConstraints::kGoogAutoGainControl, true },
  63   { MediaAudioConstraints::kGoogExperimentalAutoGainControl, true },
  64   { MediaAudioConstraints::kGoogNoiseSuppression, true },
  65   { MediaAudioConstraints::kGoogHighpassFilter, true },
  66   { MediaAudioConstraints::kGoogTypingNoiseDetection, true },
  67   { MediaAudioConstraints::kGoogExperimentalNoiseSuppression, false },
  68   { MediaAudioConstraints::kGoogBeamforming, false },
  69   { kMediaStreamAudioHotword, false },
  70 };
  71
  72 // Used to log echo quality based on delay estimates.
  73 enum DelayBasedEchoQuality {
  74   DELAY_BASED_ECHO_QUALITY_GOOD = 0,
  75   DELAY_BASED_ECHO_QUALITY_SPURIOUS,
  76   DELAY_BASED_ECHO_QUALITY_BAD,
  77   DELAY_BASED_ECHO_QUALITY_INVALID,
  78   DELAY_BASED_ECHO_QUALITY_MAX
  79 };
  80
  81 DelayBasedEchoQuality EchoDelayFrequencyToQuality(float delay_frequency) {
  82   const float kEchoDelayFrequencyLowerLimit = 0.1f;
  83   const float kEchoDelayFrequencyUpperLimit = 0.8f;
  84   // DELAY_BASED_ECHO_QUALITY_GOOD
  85   //   delay is out of bounds during at most 10 % of the time.
  86   // DELAY_BASED_ECHO_QUALITY_SPURIOUS
  87   //   delay is out of bounds 10-80 % of the time.
  88   // DELAY_BASED_ECHO_QUALITY_BAD
  89   //   delay is mostly out of bounds >= 80 % of the time.
  90   // DELAY_BASED_ECHO_QUALITY_INVALID
  91   //   delay_frequency is negative which happens if we have insufficient data.
  92   if (delay_frequency < 0)
  93     return DELAY_BASED_ECHO_QUALITY_INVALID;
  94   else if (delay_frequency <= kEchoDelayFrequencyLowerLimit)
  95     return DELAY_BASED_ECHO_QUALITY_GOOD;
  96   else if (delay_frequency < kEchoDelayFrequencyUpperLimit)
  97     return DELAY_BASED_ECHO_QUALITY_SPURIOUS;
  98   else
  99     return DELAY_BASED_ECHO_QUALITY_BAD;
 100 }
 101
 102 }  // namespace
 103
 104 // TODO(xians): Remove this method after the APM in WebRtc is deprecated.
 105 void MediaAudioConstraints::ApplyFixedAudioConstraints(
 106     RTCMediaConstraints* constraints) {
 107   for (size_t i = 0; i < arraysize(kDefaultAudioConstraints); ++i) {
 108     bool already_set_value;
 109     if (!webrtc::FindConstraint(constraints, kDefaultAudioConstraints[i].key,
 110                                 &already_set_value, NULL)) {
 111       const std::string value = kDefaultAudioConstraints[i].value ?
 112           webrtc::MediaConstraintsInterface::kValueTrue :
 113           webrtc::MediaConstraintsInterface::kValueFalse;
 114       constraints->AddOptional(kDefaultAudioConstraints[i].key, value, false);
 115     } else {
 116       DVLOG(1) << "Constraint " << kDefaultAudioConstraints[i].key
 117                << " already set to " << already_set_value;
 118     }
 119   }
 120 }
 121
 122 MediaAudioConstraints::MediaAudioConstraints(
 123     const blink::WebMediaConstraints& constraints, int effects)
 124     : constraints_(constraints),
 125       effects_(effects),
 126       default_audio_processing_constraint_value_(true) {
 127   // The default audio processing constraints are turned off when
 128   // - gUM has a specific kMediaStreamSource, which is used by tab capture
 129   //   and screen capture.
 130   // - |kEchoCancellation| is explicitly set to false.
 131   std::string value_str;
 132   bool value_bool = false;
 133   if ((GetConstraintValueAsString(constraints, kMediaStreamSource,
 134                                   &value_str)) ||
 135       (GetConstraintValueAsBoolean(constraints_, kEchoCancellation,
 136                                    &value_bool) && !value_bool)) {
 137     default_audio_processing_constraint_value_ = false;
 138   }
 139 }
 140
 141 MediaAudioConstraints::~MediaAudioConstraints() {}
 142
 143 bool MediaAudioConstraints::GetProperty(const std::string& key) const {
 144   // Return the value if the constraint is specified in |constraints|,
 145   // otherwise return the default value.
 146   bool value = false;
 147   if (!GetConstraintValueAsBoolean(constraints_, key, &value))
 148     value = GetDefaultValueForConstraint(constraints_, key);
 149
 150   return value;
 151 }
 152
 153 std::string MediaAudioConstraints::GetPropertyAsString(
 154     const std::string& key) const {
 155   std::string value;
 156   GetConstraintValueAsString(constraints_, key, &value);
 157   return value;
 158 }
 159
 160 bool MediaAudioConstraints::GetEchoCancellationProperty() const {
 161   // If platform echo canceller is enabled, disable the software AEC.
 162   if (effects_ & media::AudioParameters::ECHO_CANCELLER)
 163     return false;
 164
 165   // If |kEchoCancellation| is specified in the constraints, it will
 166   // override the value of |kGoogEchoCancellation|.
 167   bool value = false;
 168   if (GetConstraintValueAsBoolean(constraints_, kEchoCancellation, &value))
 169     return value;
 170
 171   return GetProperty(kGoogEchoCancellation);
 172 }
 173
 174 bool MediaAudioConstraints::IsValid() const {
 175   blink::WebVector<blink::WebMediaConstraint> mandatory;
 176   constraints_.getMandatoryConstraints(mandatory);
 177   for (size_t i = 0; i < mandatory.size(); ++i) {
 178     const std::string key = mandatory[i].m_name.utf8();
 179     if (key == kMediaStreamSource || key == kMediaStreamSourceId ||
 180         key == MediaStreamSource::kSourceId) {
 181       // Ignore Chrome specific Tab capture and |kSourceId| constraints.
 182       continue;
 183     }
 184
 185     bool valid = false;
 186     for (size_t j = 0; j < arraysize(kDefaultAudioConstraints); ++j) {
 187       if (key == kDefaultAudioConstraints[j].key) {
 188         bool value = false;
 189         valid = GetMandatoryConstraintValueAsBoolean(constraints_, key, &value);
 190         break;
 191       }
 192     }
 193
 194     if (!valid) {
 195       DLOG(ERROR) << "Invalid MediaStream constraint. Name: " << key;
 196       return false;
 197     }
 198   }
 199
 200   return true;
 201 }
 202
 203 bool MediaAudioConstraints::GetDefaultValueForConstraint(
 204     const blink::WebMediaConstraints& constraints,
 205     const std::string& key) const {
 206   if (!default_audio_processing_constraint_value_)
 207     return false;
 208
 209   for (size_t i = 0; i < arraysize(kDefaultAudioConstraints); ++i) {
 210     if (kDefaultAudioConstraints[i].key == key)
 211       return kDefaultAudioConstraints[i].value;
 212   }
 213
 214   return false;
 215 }
 216
 217 EchoInformation::EchoInformation()
 218     : num_chunks_(0), echo_frames_received_(false) {
 219 }
 220
 221 EchoInformation::~EchoInformation() {}
 222
 223 void EchoInformation::UpdateAecDelayStats(
 224     webrtc::EchoCancellation* echo_cancellation) {
 225   // Only start collecting stats if we know echo cancellation has measured an
 226   // echo. Otherwise we clutter the stats with for example cases where only the
 227   // microphone is used.
 228   if (!echo_frames_received_ & !echo_cancellation->stream_has_echo())
 229     return;
 230
 231   echo_frames_received_ = true;
 232   // In WebRTC, three echo delay metrics are calculated and updated every
 233   // five seconds. We use one of them, |fraction_poor_delays| to log in a UMA
 234   // histogram an Echo Cancellation quality metric. The stat in WebRTC has a
 235   // fixed aggregation window of five seconds, so we use the same query
 236   // frequency to avoid logging old values.
 237   const int kNumChunksInFiveSeconds = 500;
 238   if (!echo_cancellation->is_delay_logging_enabled() ||
 239       !echo_cancellation->is_enabled()) {
 240     return;
 241   }
 242
 243   num_chunks_++;
 244   if (num_chunks_ < kNumChunksInFiveSeconds) {
 245     return;
 246   }
 247
 248   int dummy_median = 0, dummy_std = 0;
 249   float fraction_poor_delays = 0;
 250   if (echo_cancellation->GetDelayMetrics(
 251           &dummy_median, &dummy_std, &fraction_poor_delays) ==
 252       webrtc::AudioProcessing::kNoError) {
 253     num_chunks_ = 0;
 254     // Map |fraction_poor_delays| to an Echo Cancellation quality and log in UMA
 255     // histogram. See DelayBasedEchoQuality for information on histogram
 256     // buckets.
 257     UMA_HISTOGRAM_ENUMERATION("WebRTC.AecDelayBasedQuality",
 258                               EchoDelayFrequencyToQuality(fraction_poor_delays),
 259                               DELAY_BASED_ECHO_QUALITY_MAX);
 260   }
 261 }
 262
 263 void EnableEchoCancellation(AudioProcessing* audio_processing) {
 264 #if defined(OS_ANDROID) || defined(OS_IOS)
 265   const std::string group_name =
 266       base::FieldTrialList::FindFullName("ReplaceAECMWithAEC");
 267   if (group_name.empty() ||
 268       !(group_name == "Enabled" || group_name == "DefaultEnabled")) {
 269     // Mobile devices are using AECM.
 270     int err = audio_processing->echo_control_mobile()->set_routing_mode(
 271         webrtc::EchoControlMobile::kSpeakerphone);
 272     err |= audio_processing->echo_control_mobile()->Enable(true);
 273     CHECK_EQ(err, 0);
 274     return;
 275   }
 276 #endif
 277   int err = audio_processing->echo_cancellation()->set_suppression_level(
 278       webrtc::EchoCancellation::kHighSuppression);
 279
 280   // Enable the metrics for AEC.
 281   err |= audio_processing->echo_cancellation()->enable_metrics(true);
 282   err |= audio_processing->echo_cancellation()->enable_delay_logging(true);
 283   err |= audio_processing->echo_cancellation()->Enable(true);
 284   CHECK_EQ(err, 0);
 285 }
 286
 287 void EnableNoiseSuppression(AudioProcessing* audio_processing,
 288                             webrtc::NoiseSuppression::Level ns_level) {
 289   int err = audio_processing->noise_suppression()->set_level(ns_level);
 290   err |= audio_processing->noise_suppression()->Enable(true);
 291   CHECK_EQ(err, 0);
 292 }
 293
 294 void EnableHighPassFilter(AudioProcessing* audio_processing) {
 295   CHECK_EQ(audio_processing->high_pass_filter()->Enable(true), 0);
 296 }
 297
 298 void EnableTypingDetection(AudioProcessing* audio_processing,
 299                            webrtc::TypingDetection* typing_detector) {
 300   int err = audio_processing->voice_detection()->Enable(true);
 301   err |= audio_processing->voice_detection()->set_likelihood(
 302       webrtc::VoiceDetection::kVeryLowLikelihood);
 303   CHECK_EQ(err, 0);
 304
 305   // Configure the update period to 1s (100 * 10ms) in the typing detector.
 306   typing_detector->SetParameters(0, 0, 0, 0, 0, 100);
 307 }
 308
 309 void StartEchoCancellationDump(AudioProcessing* audio_processing,
 310                                base::File aec_dump_file) {
 311   DCHECK(aec_dump_file.IsValid());
 312
 313   FILE* stream = base::FileToFILE(aec_dump_file.Pass(), "w");
 314   if (!stream) {
 315     LOG(ERROR) << "Failed to open AEC dump file";
 316     return;
 317   }
 318
 319   if (audio_processing->StartDebugRecording(stream))
 320     DLOG(ERROR) << "Fail to start AEC debug recording";
 321 }
 322
 323 void StopEchoCancellationDump(AudioProcessing* audio_processing) {
 324   if (audio_processing->StopDebugRecording())
 325     DLOG(ERROR) << "Fail to stop AEC debug recording";
 326 }
 327
 328 void EnableAutomaticGainControl(AudioProcessing* audio_processing) {
 329 #if defined(OS_ANDROID) || defined(OS_IOS)
 330   const webrtc::GainControl::Mode mode = webrtc::GainControl::kFixedDigital;
 331 #else
 332   const webrtc::GainControl::Mode mode = webrtc::GainControl::kAdaptiveAnalog;
 333 #endif
 334   int err = audio_processing->gain_control()->set_mode(mode);
 335   err |= audio_processing->gain_control()->Enable(true);
 336   CHECK_EQ(err, 0);
 337 }
 338
 339 void GetAecStats(webrtc::EchoCancellation* echo_cancellation,
 340                  webrtc::AudioProcessorInterface::AudioProcessorStats* stats) {
 341   // These values can take on valid negative values, so use the lowest possible
 342   // level as default rather than -1.
 343   stats->echo_return_loss = -100;
 344   stats->echo_return_loss_enhancement = -100;
 345
 346   // The median value can also be negative, but in practice -1 is only used to
 347   // signal insufficient data, since the resolution is limited to multiples
 348   // of 4ms.
 349   stats->echo_delay_median_ms = -1;
 350   stats->echo_delay_std_ms = -1;
 351
 352   // TODO(ajm): Re-enable this metric once we have a reliable implementation.
 353   stats->aec_quality_min = -1.0f;
 354
 355   if (!echo_cancellation->are_metrics_enabled() ||
 356       !echo_cancellation->is_delay_logging_enabled() ||
 357       !echo_cancellation->is_enabled()) {
 358     return;
 359   }
 360
 361   // TODO(ajm): we may want to use VoECallReport::GetEchoMetricsSummary
 362   // here, but it appears to be unsuitable currently. Revisit after this is
 363   // investigated: http://b/issue?id=5666755
 364   webrtc::EchoCancellation::Metrics echo_metrics;
 365   if (!echo_cancellation->GetMetrics(&echo_metrics)) {
 366     stats->echo_return_loss = echo_metrics.echo_return_loss.instant;
 367     stats->echo_return_loss_enhancement =
 368         echo_metrics.echo_return_loss_enhancement.instant;
 369   }
 370
 371   int median = 0, std = 0;
 372   float dummy = 0;
 373   if (echo_cancellation->GetDelayMetrics(&median, &std, &dummy) ==
 374       webrtc::AudioProcessing::kNoError) {
 375     stats->echo_delay_median_ms = median;
 376     stats->echo_delay_std_ms = std;
 377   }
 378 }
 379
 380 CONTENT_EXPORT std::vector<webrtc::Point> ParseArrayGeometry(
 381     const std::string& geometry_string) {
 382   const auto& tokens =
 383       base::SplitString(geometry_string, base::kWhitespaceASCII,
 384                         base::KEEP_WHITESPACE, base::SPLIT_WANT_NONEMPTY);
 385   std::vector<webrtc::Point> geometry;
 386   if (tokens.size() < 3 || tokens.size() % 3 != 0) {
 387     LOG(ERROR) << "Malformed geometry string: " << geometry_string;
 388     return geometry;
 389   }
 390
 391   std::vector<float> float_tokens;
 392   float_tokens.reserve(tokens.size());
 393   for (const auto& token : tokens) {
 394     double float_token;
 395     if (!base::StringToDouble(token, &float_token)) {
 396       LOG(ERROR) << "Unable to convert token=" << token
 397                  << " to double from geometry string: " << geometry_string;
 398       return geometry;
 399     }
 400     float_tokens.push_back(float_token);
 401   }
 402
 403   geometry.reserve(float_tokens.size() / 3);
 404   for (size_t i = 0; i < float_tokens.size(); i += 3) {
 405     geometry.push_back(webrtc::Point(float_tokens[i + 0], float_tokens[i + 1],
 406                                      float_tokens[i + 2]));
 407   }
 408
 409   return geometry;
 410 }
 411
 412 }  // namespace content