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 "base/time/time.h"
  15 #include "content/common/media/media_stream_options.h"
  16 #include "content/renderer/media/media_stream_constraints_util.h"
  17 #include "content/renderer/media/media_stream_source.h"
  18 #include "content/renderer/media/rtc_media_constraints.h"
  19 #include "media/audio/audio_parameters.h"
  20 #include "third_party/webrtc/modules/audio_processing/include/audio_processing.h"
  21 #include "third_party/webrtc/modules/audio_processing/typing_detection.h"
  22
  23 namespace content {
  24
  25 const char MediaAudioConstraints::kEchoCancellation[] = "echoCancellation";
  26 const char MediaAudioConstraints::kGoogEchoCancellation[] =
  27     "googEchoCancellation";
  28 const char MediaAudioConstraints::kGoogExperimentalEchoCancellation[] =
  29     "googEchoCancellation2";
  30 const char MediaAudioConstraints::kGoogAutoGainControl[] =
  31     "googAutoGainControl";
  32 const char MediaAudioConstraints::kGoogExperimentalAutoGainControl[] =
  33     "googAutoGainControl2";
  34 const char MediaAudioConstraints::kGoogNoiseSuppression[] =
  35     "googNoiseSuppression";
  36 const char MediaAudioConstraints::kGoogExperimentalNoiseSuppression[] =
  37     "googNoiseSuppression2";
  38 const char MediaAudioConstraints::kGoogBeamforming[] = "googBeamforming";
  39 const char MediaAudioConstraints::kGoogHighpassFilter[] = "googHighpassFilter";
  40 const char MediaAudioConstraints::kGoogTypingNoiseDetection[] =
  41     "googTypingNoiseDetection";
  42 const char MediaAudioConstraints::kGoogAudioMirroring[] = "googAudioMirroring";
  43
  44 namespace {
  45
  46 // Constant constraint keys which enables default audio constraints on
  47 // mediastreams with audio.
  48 struct {
  49   const char* key;
  50   bool value;
  51 } const kDefaultAudioConstraints[] = {
  52   { MediaAudioConstraints::kEchoCancellation, true },
  53   { MediaAudioConstraints::kGoogEchoCancellation, true },
  54 #if defined(OS_ANDROID) || defined(OS_IOS)
  55   { MediaAudioConstraints::kGoogExperimentalEchoCancellation, false },
  56 #else
  57   // Enable the extended filter mode AEC on all non-mobile platforms.
  58   { MediaAudioConstraints::kGoogExperimentalEchoCancellation, true },
  59 #endif
  60   { MediaAudioConstraints::kGoogAutoGainControl, true },
  61   { MediaAudioConstraints::kGoogExperimentalAutoGainControl, true },
  62   { MediaAudioConstraints::kGoogNoiseSuppression, true },
  63   { MediaAudioConstraints::kGoogHighpassFilter, true },
  64   { MediaAudioConstraints::kGoogTypingNoiseDetection, true },
  65   { MediaAudioConstraints::kGoogExperimentalNoiseSuppression, false },
  66   { MediaAudioConstraints::kGoogBeamforming, false },
  67 #if defined(OS_WIN)
  68   { kMediaStreamAudioDucking, true },
  69 #else
  70   { kMediaStreamAudioDucking, false },
  71 #endif
  72   { kMediaStreamAudioHotword, false },
  73 };
  74
  75 bool IsAudioProcessingConstraint(const std::string& key) {
  76   // |kMediaStreamAudioDucking| does not require audio processing.
  77   return key != kMediaStreamAudioDucking;
  78 }
  79
  80 // Used to log echo quality based on delay estimates.
  81 enum DelayBasedEchoQuality {
  82   DELAY_BASED_ECHO_QUALITY_GOOD = 0,
  83   DELAY_BASED_ECHO_QUALITY_SPURIOUS,
  84   DELAY_BASED_ECHO_QUALITY_BAD,
  85   DELAY_BASED_ECHO_QUALITY_MAX
  86 };
  87
  88 DelayBasedEchoQuality EchoDelayFrequencyToQuality(float delay_frequency) {
  89   const float kEchoDelayFrequencyLowerLimit = 0.1f;
  90   const float kEchoDelayFrequencyUpperLimit = 0.8f;
  91   // DELAY_BASED_ECHO_QUALITY_GOOD
  92   //   delay is out of bounds during at most 10 % of the time.
  93   // DELAY_BASED_ECHO_QUALITY_SPURIOUS
  94   //   delay is out of bounds 10-80 % of the time.
  95   // DELAY_BASED_ECHO_QUALITY_BAD
  96   //   delay is mostly out of bounds >= 80 % of the time.
  97   if (delay_frequency <= kEchoDelayFrequencyLowerLimit)
  98     return DELAY_BASED_ECHO_QUALITY_GOOD;
  99   else if (delay_frequency < kEchoDelayFrequencyUpperLimit)
 100     return DELAY_BASED_ECHO_QUALITY_SPURIOUS;
 101   else
 102     return DELAY_BASED_ECHO_QUALITY_BAD;
 103 }
 104
 105 }  // namespace
 106
 107 // TODO(xians): Remove this method after the APM in WebRtc is deprecated.
 108 void MediaAudioConstraints::ApplyFixedAudioConstraints(
 109     RTCMediaConstraints* constraints) {
 110   for (size_t i = 0; i < arraysize(kDefaultAudioConstraints); ++i) {
 111     bool already_set_value;
 112     if (!webrtc::FindConstraint(constraints, kDefaultAudioConstraints[i].key,
 113                                 &already_set_value, NULL)) {
 114       const std::string value = kDefaultAudioConstraints[i].value ?
 115           webrtc::MediaConstraintsInterface::kValueTrue :
 116           webrtc::MediaConstraintsInterface::kValueFalse;
 117       constraints->AddOptional(kDefaultAudioConstraints[i].key, value, false);
 118     } else {
 119       DVLOG(1) << "Constraint " << kDefaultAudioConstraints[i].key
 120                << " already set to " << already_set_value;
 121     }
 122   }
 123 }
 124
 125 MediaAudioConstraints::MediaAudioConstraints(
 126     const blink::WebMediaConstraints& constraints, int effects)
 127     : constraints_(constraints),
 128       effects_(effects),
 129       default_audio_processing_constraint_value_(true) {
 130   // The default audio processing constraints are turned off when
 131   // - gUM has a specific kMediaStreamSource, which is used by tab capture
 132   //   and screen capture.
 133   // - |kEchoCancellation| is explicitly set to false.
 134   std::string value_str;
 135   bool value_bool = false;
 136   if ((GetConstraintValueAsString(constraints, kMediaStreamSource,
 137                                   &value_str)) ||
 138       (GetConstraintValueAsBoolean(constraints_, kEchoCancellation,
 139                                    &value_bool) && !value_bool)) {
 140     default_audio_processing_constraint_value_ = false;
 141   }
 142 }
 143
 144 MediaAudioConstraints::~MediaAudioConstraints() {}
 145
 146 bool MediaAudioConstraints::GetProperty(const std::string& key) {
 147   // Return the value if the constraint is specified in |constraints|,
 148   // otherwise return the default value.
 149   bool value = false;
 150   if (!GetConstraintValueAsBoolean(constraints_, key, &value))
 151     value = GetDefaultValueForConstraint(constraints_, key);
 152
 153   return value;
 154 }
 155
 156 bool MediaAudioConstraints::GetEchoCancellationProperty() {
 157   // If platform echo canceller is enabled, disable the software AEC.
 158   if (effects_ & media::AudioParameters::ECHO_CANCELLER)
 159     return false;
 160
 161   // If |kEchoCancellation| is specified in the constraints, it will
 162   // override the value of |kGoogEchoCancellation|.
 163   bool value = false;
 164   if (GetConstraintValueAsBoolean(constraints_, kEchoCancellation, &value))
 165     return value;
 166
 167   return GetProperty(kGoogEchoCancellation);
 168 }
 169
 170 bool MediaAudioConstraints::IsValid() {
 171   blink::WebVector<blink::WebMediaConstraint> mandatory;
 172   constraints_.getMandatoryConstraints(mandatory);
 173   for (size_t i = 0; i < mandatory.size(); ++i) {
 174     const std::string key = mandatory[i].m_name.utf8();
 175     if (key == kMediaStreamSource || key == kMediaStreamSourceId ||
 176         key == MediaStreamSource::kSourceId) {
 177       // Ignore Chrome specific Tab capture and |kSourceId| constraints.
 178       continue;
 179     }
 180
 181     bool valid = false;
 182     for (size_t j = 0; j < arraysize(kDefaultAudioConstraints); ++j) {
 183       if (key == kDefaultAudioConstraints[j].key) {
 184         bool value = false;
 185         valid = GetMandatoryConstraintValueAsBoolean(constraints_, key, &value);
 186         break;
 187       }
 188     }
 189
 190     if (!valid) {
 191       DLOG(ERROR) << "Invalid MediaStream constraint. Name: " << key;
 192       return false;
 193     }
 194   }
 195
 196   return true;
 197 }
 198
 199 bool MediaAudioConstraints::GetDefaultValueForConstraint(
 200     const blink::WebMediaConstraints& constraints, const std::string& key) {
 201   // |kMediaStreamAudioDucking| is not restricted by
 202   // |default_audio_processing_constraint_value_| since it does not require
 203   // audio processing.
 204   if (!default_audio_processing_constraint_value_ &&
 205       IsAudioProcessingConstraint(key))
 206     return false;
 207
 208   for (size_t i = 0; i < arraysize(kDefaultAudioConstraints); ++i) {
 209     if (kDefaultAudioConstraints[i].key == key)
 210       return kDefaultAudioConstraints[i].value;
 211   }
 212
 213   return false;
 214 }
 215
 216 EchoInformation::EchoInformation()
 217     : echo_poor_delay_counts_(0),
 218       echo_total_delay_counts_(0),
 219       last_log_time_(base::TimeTicks::Now()) {}
 220
 221 EchoInformation::~EchoInformation() {}
 222
 223 void EchoInformation::UpdateAecDelayStats(int delay) {
 224   // One way to get an indication of how well the echo cancellation performs is
 225   // to compare the, by AEC, estimated delay with the AEC filter length.
 226   // |kMaxAecFilterLengthMs| is the maximum delay we can allow before we
 227   // consider the AEC to fail. This value should not be larger than the filter
 228   // length used inside AEC. This is for now set to match the extended filter
 229   // mode which is turned on for all platforms.
 230   const int kMaxAecFilterLengthMs = 128;
 231   if ((delay < -2) || (delay > kMaxAecFilterLengthMs)) {
 232     // The |delay| is out of bounds which indicates that the echo cancellation
 233     // filter can not handle the echo. Hence, we have a potential full echo
 234     // case. |delay| values {-1, -2} are reserved for errors.
 235     ++echo_poor_delay_counts_;
 236   }
 237   ++echo_total_delay_counts_;
 238   LogAecDelayStats();
 239 }
 240
 241 void EchoInformation::LogAecDelayStats() {
 242   // We update the UMA statistics every 5 seconds.
 243   const int kTimeBetweenLogsInSeconds = 5;
 244   const base::TimeDelta time_since_last_log =
 245       base::TimeTicks::Now() - last_log_time_;
 246   if (time_since_last_log.InSeconds() < kTimeBetweenLogsInSeconds)
 247     return;
 248
 249   // Calculate how frequent the AEC delay was out of bounds since last time we
 250   // updated UMA histograms. Then store the result into one of three histogram
 251   // buckets; see DelayBasedEchoQuality.
 252   float poor_delay_frequency = 0.f;
 253   if (echo_total_delay_counts_ > 0) {
 254     poor_delay_frequency = static_cast<float>(echo_poor_delay_counts_) /
 255         static_cast<float>(echo_total_delay_counts_);
 256     UMA_HISTOGRAM_ENUMERATION("WebRTC.AecDelayBasedQuality",
 257                               EchoDelayFrequencyToQuality(poor_delay_frequency),
 258                               DELAY_BASED_ECHO_QUALITY_MAX);
 259   }
 260   echo_poor_delay_counts_ = 0;
 261   echo_total_delay_counts_ = 0;
 262   last_log_time_ = base::TimeTicks::Now();
 263 }
 264
 265 void EnableEchoCancellation(AudioProcessing* audio_processing) {
 266 #if defined(OS_ANDROID) || defined(OS_IOS)
 267   const std::string group_name =
 268       base::FieldTrialList::FindFullName("ReplaceAECMWithAEC");
 269   if (group_name.empty() ||
 270       !(group_name == "Enabled" || group_name == "DefaultEnabled")) {
 271     // Mobile devices are using AECM.
 272     int err = audio_processing->echo_control_mobile()->set_routing_mode(
 273         webrtc::EchoControlMobile::kSpeakerphone);
 274     err |= audio_processing->echo_control_mobile()->Enable(true);
 275     CHECK_EQ(err, 0);
 276     return;
 277   }
 278 #endif
 279   int err = audio_processing->echo_cancellation()->set_suppression_level(
 280       webrtc::EchoCancellation::kHighSuppression);
 281
 282   // Enable the metrics for AEC.
 283   err |= audio_processing->echo_cancellation()->enable_metrics(true);
 284   err |= audio_processing->echo_cancellation()->enable_delay_logging(true);
 285   err |= audio_processing->echo_cancellation()->Enable(true);
 286   CHECK_EQ(err, 0);
 287 }
 288
 289 void EnableNoiseSuppression(AudioProcessing* audio_processing) {
 290   int err = audio_processing->noise_suppression()->set_level(
 291       webrtc::NoiseSuppression::kHigh);
 292   err |= audio_processing->noise_suppression()->Enable(true);
 293   CHECK_EQ(err, 0);
 294 }
 295
 296 void EnableHighPassFilter(AudioProcessing* audio_processing) {
 297   CHECK_EQ(audio_processing->high_pass_filter()->Enable(true), 0);
 298 }
 299
 300 void EnableTypingDetection(AudioProcessing* audio_processing,
 301                            webrtc::TypingDetection* typing_detector) {
 302   int err = audio_processing->voice_detection()->Enable(true);
 303   err |= audio_processing->voice_detection()->set_likelihood(
 304       webrtc::VoiceDetection::kVeryLowLikelihood);
 305   CHECK_EQ(err, 0);
 306
 307   // Configure the update period to 1s (100 * 10ms) in the typing detector.
 308   typing_detector->SetParameters(0, 0, 0, 0, 0, 100);
 309 }
 310
 311 void StartEchoCancellationDump(AudioProcessing* audio_processing,
 312                                base::File aec_dump_file) {
 313   DCHECK(aec_dump_file.IsValid());
 314
 315   FILE* stream = base::FileToFILE(aec_dump_file.Pass(), "w");
 316   if (!stream) {
 317     LOG(ERROR) << "Failed to open AEC dump file";
 318     return;
 319   }
 320
 321   if (audio_processing->StartDebugRecording(stream))
 322     DLOG(ERROR) << "Fail to start AEC debug recording";
 323 }
 324
 325 void StopEchoCancellationDump(AudioProcessing* audio_processing) {
 326   if (audio_processing->StopDebugRecording())
 327     DLOG(ERROR) << "Fail to stop AEC debug recording";
 328 }
 329
 330 void EnableAutomaticGainControl(AudioProcessing* audio_processing) {
 331 #if defined(OS_ANDROID) || defined(OS_IOS)
 332   const webrtc::GainControl::Mode mode = webrtc::GainControl::kFixedDigital;
 333 #else
 334   const webrtc::GainControl::Mode mode = webrtc::GainControl::kAdaptiveAnalog;
 335 #endif
 336   int err = audio_processing->gain_control()->set_mode(mode);
 337   err |= audio_processing->gain_control()->Enable(true);
 338   CHECK_EQ(err, 0);
 339 }
 340
 341 void GetAecStats(AudioProcessing* audio_processing,
 342                  webrtc::AudioProcessorInterface::AudioProcessorStats* stats) {
 343   // These values can take on valid negative values, so use the lowest possible
 344   // level as default rather than -1.
 345   stats->echo_return_loss = -100;
 346   stats->echo_return_loss_enhancement = -100;
 347
 348   // These values can also be negative, but in practice -1 is only used to
 349   // signal insufficient data, since the resolution is limited to multiples
 350   // of 4ms.
 351   stats->echo_delay_median_ms = -1;
 352   stats->echo_delay_std_ms = -1;
 353
 354   // TODO(ajm): Re-enable this metric once we have a reliable implementation.
 355   stats->aec_quality_min = -1.0f;
 356
 357   if (!audio_processing->echo_cancellation()->are_metrics_enabled() ||
 358       !audio_processing->echo_cancellation()->is_delay_logging_enabled() ||
 359       !audio_processing->echo_cancellation()->is_enabled()) {
 360     return;
 361   }
 362
 363   // TODO(ajm): we may want to use VoECallReport::GetEchoMetricsSummary
 364   // here, but it appears to be unsuitable currently. Revisit after this is
 365   // investigated: http://b/issue?id=5666755
 366   webrtc::EchoCancellation::Metrics echo_metrics;
 367   if (!audio_processing->echo_cancellation()->GetMetrics(&echo_metrics)) {
 368     stats->echo_return_loss = echo_metrics.echo_return_loss.instant;
 369     stats->echo_return_loss_enhancement =
 370         echo_metrics.echo_return_loss_enhancement.instant;
 371   }
 372
 373   int median = 0, std = 0;
 374   if (!audio_processing->echo_cancellation()->GetDelayMetrics(&median, &std)) {
 375     stats->echo_delay_median_ms = median;
 376     stats->echo_delay_std_ms = std;
 377   }
 378 }
 379
 380 }  // namespace content