src/content/browser/speech/speech_recognition_manager_impl.cc

   1 // Copyright (c) 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/browser/speech/speech_recognition_manager_impl.h"
   6
   7 #include "base/bind.h"
   8 #include "content/browser/browser_main_loop.h"
   9 #include "content/browser/renderer_host/media/media_stream_manager.h"
  10 #include "content/browser/renderer_host/media/media_stream_ui_proxy.h"
  11 #include "content/browser/speech/google_one_shot_remote_engine.h"
  12 #include "content/browser/speech/google_streaming_remote_engine.h"
  13 #include "content/browser/speech/speech_recognition_engine.h"
  14 #include "content/browser/speech/speech_recognizer_impl.h"
  15 #include "content/public/browser/browser_thread.h"
  16 #include "content/public/browser/content_browser_client.h"
  17 #include "content/public/browser/resource_context.h"
  18 #include "content/public/browser/speech_recognition_event_listener.h"
  19 #include "content/public/browser/speech_recognition_manager_delegate.h"
  20 #include "content/public/browser/speech_recognition_session_config.h"
  21 #include "content/public/browser/speech_recognition_session_context.h"
  22 #include "content/public/common/speech_recognition_error.h"
  23 #include "content/public/common/speech_recognition_result.h"
  24 #include "media/audio/audio_manager.h"
  25 #include "media/audio/audio_manager_base.h"
  26
  27 #if defined(OS_ANDROID)
  28 #include "content/browser/speech/speech_recognizer_impl_android.h"
  29 #endif
  30
  31 using base::Callback;
  32
  33 namespace content {
  34
  35 SpeechRecognitionManager* SpeechRecognitionManager::manager_for_tests_;
  36
  37 namespace {
  38
  39 SpeechRecognitionManagerImpl* g_speech_recognition_manager_impl;
  40
  41 void ShowAudioInputSettingsOnFileThread(media::AudioManager* audio_manager) {
  42   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::FILE));
  43   audio_manager->ShowAudioInputSettings();
  44 }
  45
  46 }  // namespace
  47
  48 SpeechRecognitionManager* SpeechRecognitionManager::GetInstance() {
  49   if (manager_for_tests_)
  50     return manager_for_tests_;
  51   return SpeechRecognitionManagerImpl::GetInstance();
  52 }
  53
  54 void SpeechRecognitionManager::SetManagerForTesting(
  55     SpeechRecognitionManager* manager) {
  56   manager_for_tests_ = manager;
  57 }
  58
  59 SpeechRecognitionManagerImpl* SpeechRecognitionManagerImpl::GetInstance() {
  60   return g_speech_recognition_manager_impl;
  61 }
  62
  63 SpeechRecognitionManagerImpl::SpeechRecognitionManagerImpl(
  64       media::AudioManager* audio_manager,
  65       MediaStreamManager* media_stream_manager)
  66     : audio_manager_(audio_manager),
  67       media_stream_manager_(media_stream_manager),
  68       primary_session_id_(kSessionIDInvalid),
  69       last_session_id_(kSessionIDInvalid),
  70       is_dispatching_event_(false),
  71       delegate_(GetContentClient()->browser()->
  72                     GetSpeechRecognitionManagerDelegate()),
  73       weak_factory_(this) {
  74   DCHECK(!g_speech_recognition_manager_impl);
  75   g_speech_recognition_manager_impl = this;
  76 }
  77
  78 SpeechRecognitionManagerImpl::~SpeechRecognitionManagerImpl() {
  79   DCHECK(g_speech_recognition_manager_impl);
  80   g_speech_recognition_manager_impl = NULL;
  81
  82   for (SessionsTable::iterator it = sessions_.begin(); it != sessions_.end();
  83        ++it) {
  84     // MediaStreamUIProxy must be deleted on the IO thread.
  85     BrowserThread::DeleteSoon(BrowserThread::IO, FROM_HERE,
  86                               it->second->ui.release());
  87     delete it->second;
  88   }
  89   sessions_.clear();
  90 }
  91
  92 int SpeechRecognitionManagerImpl::CreateSession(
  93     const SpeechRecognitionSessionConfig& config) {
  94   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
  95
  96   const int session_id = GetNextSessionID();
  97   DCHECK(!SessionExists(session_id));
  98   // Set-up the new session.
  99   Session* session = new Session();
 100   sessions_[session_id] = session;
 101   session->id = session_id;
 102   session->config = config;
 103   session->context = config.initial_context;
 104
 105   std::string hardware_info;
 106   bool can_report_metrics = false;
 107   if (delegate_)
 108     delegate_->GetDiagnosticInformation(&can_report_metrics, &hardware_info);
 109
 110   // The legacy api cannot use continuous mode.
 111   DCHECK(!config.is_legacy_api || !config.continuous);
 112
 113 #if !defined(OS_ANDROID)
 114   // A SpeechRecognitionEngine (and corresponding Config) is required only
 115   // when using SpeechRecognizerImpl, which performs the audio capture and
 116   // endpointing in the browser. This is not the case of Android where, not
 117   // only the speech recognition, but also the audio capture and endpointing
 118   // activities performed outside of the browser (delegated via JNI to the
 119   // Android API implementation).
 120
 121   SpeechRecognitionEngineConfig remote_engine_config;
 122   remote_engine_config.language = config.language;
 123   remote_engine_config.grammars = config.grammars;
 124   remote_engine_config.audio_sample_rate =
 125       SpeechRecognizerImpl::kAudioSampleRate;
 126   remote_engine_config.audio_num_bits_per_sample =
 127       SpeechRecognizerImpl::kNumBitsPerAudioSample;
 128   remote_engine_config.filter_profanities = config.filter_profanities;
 129   remote_engine_config.continuous = config.continuous;
 130   remote_engine_config.interim_results = config.interim_results;
 131   remote_engine_config.max_hypotheses = config.max_hypotheses;
 132   remote_engine_config.hardware_info = hardware_info;
 133   remote_engine_config.origin_url =
 134       can_report_metrics ? config.origin_url : std::string();
 135
 136   SpeechRecognitionEngine* google_remote_engine;
 137   if (config.is_legacy_api) {
 138     google_remote_engine =
 139         new GoogleOneShotRemoteEngine(config.url_request_context_getter.get());
 140   } else {
 141     google_remote_engine = new GoogleStreamingRemoteEngine(
 142         config.url_request_context_getter.get());
 143   }
 144
 145   google_remote_engine->SetConfig(remote_engine_config);
 146
 147   session->recognizer = new SpeechRecognizerImpl(
 148       this,
 149       session_id,
 150       config.continuous,
 151       config.interim_results,
 152       google_remote_engine);
 153 #else
 154   session->recognizer = new SpeechRecognizerImplAndroid(this, session_id);
 155 #endif
 156   return session_id;
 157 }
 158
 159 void SpeechRecognitionManagerImpl::StartSession(int session_id) {
 160   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 161   if (!SessionExists(session_id))
 162     return;
 163
 164   // If there is another active session, abort that.
 165   if (primary_session_id_ != kSessionIDInvalid &&
 166       primary_session_id_ != session_id) {
 167     AbortSession(primary_session_id_);
 168   }
 169
 170   primary_session_id_ = session_id;
 171
 172   if (delegate_) {
 173     delegate_->CheckRecognitionIsAllowed(
 174         session_id,
 175         base::Bind(&SpeechRecognitionManagerImpl::RecognitionAllowedCallback,
 176                    weak_factory_.GetWeakPtr(),
 177                    session_id));
 178   }
 179 }
 180
 181 void SpeechRecognitionManagerImpl::RecognitionAllowedCallback(int session_id,
 182                                                               bool ask_user,
 183                                                               bool is_allowed) {
 184   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 185   if (!SessionExists(session_id))
 186     return;
 187
 188   SessionsTable::iterator iter = sessions_.find(session_id);
 189   DCHECK(iter != sessions_.end());
 190   Session* session = iter->second;
 191
 192   if (session->abort_requested)
 193     return;
 194
 195   if (ask_user) {
 196     SpeechRecognitionSessionContext& context = session->context;
 197     context.label = media_stream_manager_->MakeMediaAccessRequest(
 198         context.render_process_id,
 199         context.render_frame_id,
 200         context.request_id,
 201         StreamOptions(true, false),
 202         GURL(context.context_name),
 203         base::Bind(
 204             &SpeechRecognitionManagerImpl::MediaRequestPermissionCallback,
 205             weak_factory_.GetWeakPtr(), session_id));
 206     return;
 207   }
 208
 209   if (is_allowed) {
 210     base::MessageLoop::current()->PostTask(
 211         FROM_HERE,
 212         base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
 213                    weak_factory_.GetWeakPtr(),
 214                    session_id,
 215                    EVENT_START));
 216   } else {
 217     OnRecognitionError(session_id, SpeechRecognitionError(
 218         SPEECH_RECOGNITION_ERROR_NOT_ALLOWED));
 219     base::MessageLoop::current()->PostTask(
 220         FROM_HERE,
 221         base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
 222                    weak_factory_.GetWeakPtr(),
 223                    session_id,
 224                    EVENT_ABORT));
 225   }
 226 }
 227
 228 void SpeechRecognitionManagerImpl::MediaRequestPermissionCallback(
 229     int session_id,
 230     const MediaStreamDevices& devices,
 231     scoped_ptr<MediaStreamUIProxy> stream_ui) {
 232   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 233
 234   SessionsTable::iterator iter = sessions_.find(session_id);
 235   if (iter == sessions_.end())
 236     return;
 237
 238   bool is_allowed = !devices.empty();
 239   if (is_allowed) {
 240     // Copy the approved devices array to the context for UI indication.
 241     iter->second->context.devices = devices;
 242
 243     // Save the UI object.
 244     iter->second->ui = stream_ui.Pass();
 245   }
 246
 247   // Clear the label to indicate the request has been done.
 248   iter->second->context.label.clear();
 249
 250   // Notify the recognition about the request result.
 251   RecognitionAllowedCallback(iter->first, false, is_allowed);
 252 }
 253
 254 void SpeechRecognitionManagerImpl::AbortSession(int session_id) {
 255   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 256   if (!SessionExists(session_id))
 257     return;
 258
 259   SessionsTable::iterator iter = sessions_.find(session_id);
 260   iter->second->ui.reset();
 261
 262   if (iter->second->abort_requested)
 263     return;
 264
 265   iter->second->abort_requested = true;
 266
 267   base::MessageLoop::current()->PostTask(
 268       FROM_HERE,
 269       base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
 270                  weak_factory_.GetWeakPtr(),
 271                  session_id,
 272                  EVENT_ABORT));
 273 }
 274
 275 void SpeechRecognitionManagerImpl::StopAudioCaptureForSession(int session_id) {
 276   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 277   if (!SessionExists(session_id))
 278     return;
 279
 280   SessionsTable::iterator iter = sessions_.find(session_id);
 281   iter->second->ui.reset();
 282
 283   base::MessageLoop::current()->PostTask(
 284       FROM_HERE,
 285       base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
 286                  weak_factory_.GetWeakPtr(),
 287                  session_id,
 288                  EVENT_STOP_CAPTURE));
 289 }
 290
 291 // Here begins the SpeechRecognitionEventListener interface implementation,
 292 // which will simply relay the events to the proper listener registered for the
 293 // particular session and to the catch-all listener provided by the delegate
 294 // (if any).
 295
 296 void SpeechRecognitionManagerImpl::OnRecognitionStart(int session_id) {
 297   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 298   if (!SessionExists(session_id))
 299     return;
 300
 301   SessionsTable::iterator iter = sessions_.find(session_id);
 302   if (iter->second->ui) {
 303     // Notify the UI that the devices are being used.
 304     iter->second->ui->OnStarted(base::Closure(),
 305                                 MediaStreamUIProxy::WindowIdCallback());
 306   }
 307
 308   DCHECK_EQ(primary_session_id_, session_id);
 309   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 310     delegate_listener->OnRecognitionStart(session_id);
 311   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 312     listener->OnRecognitionStart(session_id);
 313 }
 314
 315 void SpeechRecognitionManagerImpl::OnAudioStart(int session_id) {
 316   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 317   if (!SessionExists(session_id))
 318     return;
 319
 320   DCHECK_EQ(primary_session_id_, session_id);
 321   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 322     delegate_listener->OnAudioStart(session_id);
 323   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 324     listener->OnAudioStart(session_id);
 325 }
 326
 327 void SpeechRecognitionManagerImpl::OnEnvironmentEstimationComplete(
 328     int session_id) {
 329   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 330   if (!SessionExists(session_id))
 331     return;
 332
 333   DCHECK_EQ(primary_session_id_, session_id);
 334   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 335     delegate_listener->OnEnvironmentEstimationComplete(session_id);
 336   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 337     listener->OnEnvironmentEstimationComplete(session_id);
 338 }
 339
 340 void SpeechRecognitionManagerImpl::OnSoundStart(int session_id) {
 341   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 342   if (!SessionExists(session_id))
 343     return;
 344
 345   DCHECK_EQ(primary_session_id_, session_id);
 346   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 347     delegate_listener->OnSoundStart(session_id);
 348   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 349     listener->OnSoundStart(session_id);
 350 }
 351
 352 void SpeechRecognitionManagerImpl::OnSoundEnd(int session_id) {
 353   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 354   if (!SessionExists(session_id))
 355     return;
 356
 357   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 358     delegate_listener->OnSoundEnd(session_id);
 359   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 360     listener->OnSoundEnd(session_id);
 361 }
 362
 363 void SpeechRecognitionManagerImpl::OnAudioEnd(int session_id) {
 364   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 365   if (!SessionExists(session_id))
 366     return;
 367
 368   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 369     delegate_listener->OnAudioEnd(session_id);
 370   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 371     listener->OnAudioEnd(session_id);
 372   base::MessageLoop::current()->PostTask(
 373       FROM_HERE,
 374       base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
 375                  weak_factory_.GetWeakPtr(),
 376                  session_id,
 377                  EVENT_AUDIO_ENDED));
 378 }
 379
 380 void SpeechRecognitionManagerImpl::OnRecognitionResults(
 381     int session_id, const SpeechRecognitionResults& results) {
 382   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 383   if (!SessionExists(session_id))
 384     return;
 385
 386   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 387     delegate_listener->OnRecognitionResults(session_id, results);
 388   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 389     listener->OnRecognitionResults(session_id, results);
 390 }
 391
 392 void SpeechRecognitionManagerImpl::OnRecognitionError(
 393     int session_id, const SpeechRecognitionError& error) {
 394   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 395   if (!SessionExists(session_id))
 396     return;
 397
 398   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 399     delegate_listener->OnRecognitionError(session_id, error);
 400   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 401     listener->OnRecognitionError(session_id, error);
 402 }
 403
 404 void SpeechRecognitionManagerImpl::OnAudioLevelsChange(
 405     int session_id, float volume, float noise_volume) {
 406   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 407   if (!SessionExists(session_id))
 408     return;
 409
 410   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 411     delegate_listener->OnAudioLevelsChange(session_id, volume, noise_volume);
 412   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 413     listener->OnAudioLevelsChange(session_id, volume, noise_volume);
 414 }
 415
 416 void SpeechRecognitionManagerImpl::OnRecognitionEnd(int session_id) {
 417   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 418   if (!SessionExists(session_id))
 419     return;
 420
 421   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
 422     delegate_listener->OnRecognitionEnd(session_id);
 423   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
 424     listener->OnRecognitionEnd(session_id);
 425   base::MessageLoop::current()->PostTask(
 426       FROM_HERE,
 427       base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
 428                  weak_factory_.GetWeakPtr(),
 429                  session_id,
 430                  EVENT_RECOGNITION_ENDED));
 431 }
 432
 433 int SpeechRecognitionManagerImpl::GetSession(
 434     int render_process_id, int render_view_id, int request_id) const {
 435   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 436   SessionsTable::const_iterator iter;
 437   for(iter = sessions_.begin(); iter != sessions_.end(); ++iter) {
 438     const int session_id = iter->first;
 439     const SpeechRecognitionSessionContext& context = iter->second->context;
 440     if (context.render_process_id == render_process_id &&
 441         context.render_view_id == render_view_id &&
 442         context.request_id == request_id) {
 443       return session_id;
 444     }
 445   }
 446   return kSessionIDInvalid;
 447 }
 448
 449 SpeechRecognitionSessionContext
 450 SpeechRecognitionManagerImpl::GetSessionContext(int session_id) const {
 451   return GetSession(session_id)->context;
 452 }
 453
 454 void SpeechRecognitionManagerImpl::AbortAllSessionsForRenderProcess(
 455     int render_process_id) {
 456   // This method gracefully destroys sessions for the listener. However, since
 457   // the listener itself is likely to be destroyed after this call, we avoid
 458   // dispatching further events to it, marking the |listener_is_active| flag.
 459   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 460   for (SessionsTable::iterator it = sessions_.begin(); it != sessions_.end();
 461        ++it) {
 462     Session* session = it->second;
 463     if (session->context.render_process_id == render_process_id) {
 464       AbortSession(session->id);
 465       session->listener_is_active = false;
 466     }
 467   }
 468 }
 469
 470 void SpeechRecognitionManagerImpl::AbortAllSessionsForRenderView(
 471     int render_process_id,
 472     int render_view_id) {
 473   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 474   for (SessionsTable::iterator it = sessions_.begin(); it != sessions_.end();
 475        ++it) {
 476     Session* session = it->second;
 477     if (session->context.render_process_id == render_process_id &&
 478         session->context.render_view_id == render_view_id) {
 479       AbortSession(session->id);
 480     }
 481   }
 482 }
 483
 484 // -----------------------  Core FSM implementation ---------------------------
 485 void SpeechRecognitionManagerImpl::DispatchEvent(int session_id,
 486                                                  FSMEvent event) {
 487   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 488
 489   // There are some corner cases in which the session might be deleted (due to
 490   // an EndRecognition event) between a request (e.g. Abort) and its dispatch.
 491   if (!SessionExists(session_id))
 492     return;
 493
 494   Session* session = GetSession(session_id);
 495   FSMState session_state = GetSessionState(session_id);
 496   DCHECK_LE(session_state, SESSION_STATE_MAX_VALUE);
 497   DCHECK_LE(event, EVENT_MAX_VALUE);
 498
 499   // Event dispatching must be sequential, otherwise it will break all the rules
 500   // and the assumptions of the finite state automata model.
 501   DCHECK(!is_dispatching_event_);
 502   is_dispatching_event_ = true;
 503   ExecuteTransitionAndGetNextState(session, session_state, event);
 504   is_dispatching_event_ = false;
 505 }
 506
 507 // This FSM handles the evolution of each session, from the viewpoint of the
 508 // interaction with the user (that may be either the browser end-user which
 509 // interacts with UI bubbles, or JS developer intracting with JS methods).
 510 // All the events received by the SpeechRecognizer instances (one for each
 511 // session) are always routed to the SpeechRecognitionEventListener(s)
 512 // regardless the choices taken in this FSM.
 513 void SpeechRecognitionManagerImpl::ExecuteTransitionAndGetNextState(
 514     Session* session, FSMState session_state, FSMEvent event) {
 515   // Note: since we're not tracking the state of the recognizer object, rather
 516   // we're directly retrieving it (through GetSessionState), we see its events
 517   // (that are AUDIO_ENDED and RECOGNITION_ENDED) after its state evolution
 518   // (e.g., when we receive the AUDIO_ENDED event, the recognizer has just
 519   // completed the transition from CAPTURING_AUDIO to WAITING_FOR_RESULT, thus
 520   // we perceive the AUDIO_ENDED event in WAITING_FOR_RESULT).
 521   // This makes the code below a bit tricky but avoids a lot of code for
 522   // tracking and reconstructing asynchronously the state of the recognizer.
 523   switch (session_state) {
 524     case SESSION_STATE_IDLE:
 525       switch (event) {
 526         case EVENT_START:
 527           return SessionStart(*session);
 528         case EVENT_ABORT:
 529           return SessionAbort(*session);
 530         case EVENT_RECOGNITION_ENDED:
 531           return SessionDelete(session);
 532         case EVENT_STOP_CAPTURE:
 533           return SessionStopAudioCapture(*session);
 534         case EVENT_AUDIO_ENDED:
 535           return;
 536       }
 537       break;
 538     case SESSION_STATE_CAPTURING_AUDIO:
 539       switch (event) {
 540         case EVENT_STOP_CAPTURE:
 541           return SessionStopAudioCapture(*session);
 542         case EVENT_ABORT:
 543           return SessionAbort(*session);
 544         case EVENT_START:
 545           return;
 546         case EVENT_AUDIO_ENDED:
 547         case EVENT_RECOGNITION_ENDED:
 548           return NotFeasible(*session, event);
 549       }
 550       break;
 551     case SESSION_STATE_WAITING_FOR_RESULT:
 552       switch (event) {
 553         case EVENT_ABORT:
 554           return SessionAbort(*session);
 555         case EVENT_AUDIO_ENDED:
 556           return ResetCapturingSessionId(*session);
 557         case EVENT_START:
 558         case EVENT_STOP_CAPTURE:
 559           return;
 560         case EVENT_RECOGNITION_ENDED:
 561           return NotFeasible(*session, event);
 562       }
 563       break;
 564   }
 565   return NotFeasible(*session, event);
 566 }
 567
 568 SpeechRecognitionManagerImpl::FSMState
 569 SpeechRecognitionManagerImpl::GetSessionState(int session_id) const {
 570   Session* session = GetSession(session_id);
 571   if (!session->recognizer.get() || !session->recognizer->IsActive())
 572     return SESSION_STATE_IDLE;
 573   if (session->recognizer->IsCapturingAudio())
 574     return SESSION_STATE_CAPTURING_AUDIO;
 575   return SESSION_STATE_WAITING_FOR_RESULT;
 576 }
 577
 578 // ----------- Contract for all the FSM evolution functions below -------------
 579 //  - Are guaranteed to be executed in the IO thread;
 580 //  - Are guaranteed to be not reentrant (themselves and each other);
 581
 582 void SpeechRecognitionManagerImpl::SessionStart(const Session& session) {
 583   DCHECK_EQ(primary_session_id_, session.id);
 584   const MediaStreamDevices& devices = session.context.devices;
 585   std::string device_id;
 586   if (devices.empty()) {
 587     // From the ask_user=false path, use the default device.
 588     // TODO(xians): Abort the session after we do not need to support this path
 589     // anymore.
 590     device_id = media::AudioManagerBase::kDefaultDeviceId;
 591   } else {
 592     // From the ask_user=true path, use the selected device.
 593     DCHECK_EQ(1u, devices.size());
 594     DCHECK_EQ(MEDIA_DEVICE_AUDIO_CAPTURE, devices.front().type);
 595     device_id = devices.front().id;
 596   }
 597
 598   session.recognizer->StartRecognition(device_id);
 599 }
 600
 601 void SpeechRecognitionManagerImpl::SessionAbort(const Session& session) {
 602   if (primary_session_id_ == session.id)
 603     primary_session_id_ = kSessionIDInvalid;
 604   DCHECK(session.recognizer.get());
 605   session.recognizer->AbortRecognition();
 606 }
 607
 608 void SpeechRecognitionManagerImpl::SessionStopAudioCapture(
 609     const Session& session) {
 610   DCHECK(session.recognizer.get());
 611   session.recognizer->StopAudioCapture();
 612 }
 613
 614 void SpeechRecognitionManagerImpl::ResetCapturingSessionId(
 615     const Session& session) {
 616   DCHECK_EQ(primary_session_id_, session.id);
 617   primary_session_id_ = kSessionIDInvalid;
 618 }
 619
 620 void SpeechRecognitionManagerImpl::SessionDelete(Session* session) {
 621   DCHECK(session->recognizer.get() == NULL || !session->recognizer->IsActive());
 622   if (primary_session_id_ == session->id)
 623     primary_session_id_ = kSessionIDInvalid;
 624   if (!session->context.label.empty())
 625     media_stream_manager_->CancelRequest(session->context.label);
 626   sessions_.erase(session->id);
 627   delete session;
 628 }
 629
 630 void SpeechRecognitionManagerImpl::NotFeasible(const Session& session,
 631                                                FSMEvent event) {
 632   NOTREACHED() << "Unfeasible event " << event
 633                << " in state " << GetSessionState(session.id)
 634                << " for session " << session.id;
 635 }
 636
 637 int SpeechRecognitionManagerImpl::GetNextSessionID() {
 638   ++last_session_id_;
 639   // Deal with wrapping of last_session_id_. (How civilized).
 640   if (last_session_id_ <= 0)
 641     last_session_id_ = 1;
 642   return last_session_id_;
 643 }
 644
 645 bool SpeechRecognitionManagerImpl::SessionExists(int session_id) const {
 646   return sessions_.find(session_id) != sessions_.end();
 647 }
 648
 649 SpeechRecognitionManagerImpl::Session*
 650 SpeechRecognitionManagerImpl::GetSession(int session_id) const {
 651   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 652   SessionsTable::const_iterator iter = sessions_.find(session_id);
 653   DCHECK(iter != sessions_.end());
 654   return iter->second;
 655 }
 656
 657 SpeechRecognitionEventListener* SpeechRecognitionManagerImpl::GetListener(
 658     int session_id) const {
 659   Session* session = GetSession(session_id);
 660   if (session->listener_is_active && session->config.event_listener)
 661     return session->config.event_listener.get();
 662   return NULL;
 663 }
 664
 665 SpeechRecognitionEventListener*
 666 SpeechRecognitionManagerImpl::GetDelegateListener() const {
 667   return delegate_.get() ? delegate_->GetEventListener() : NULL;
 668 }
 669
 670 const SpeechRecognitionSessionConfig&
 671 SpeechRecognitionManagerImpl::GetSessionConfig(int session_id) const {
 672   return GetSession(session_id)->config;
 673 }
 674
 675 bool SpeechRecognitionManagerImpl::HasAudioInputDevices() {
 676   return audio_manager_->HasAudioInputDevices();
 677 }
 678
 679 base::string16 SpeechRecognitionManagerImpl::GetAudioInputDeviceModel() {
 680   return audio_manager_->GetAudioInputDeviceModel();
 681 }
 682
 683 void SpeechRecognitionManagerImpl::ShowAudioInputSettings() {
 684   // Since AudioManager::ShowAudioInputSettings can potentially launch external
 685   // processes, do that in the FILE thread to not block the calling threads.
 686   BrowserThread::PostTask(BrowserThread::FILE, FROM_HERE,
 687                           base::Bind(&ShowAudioInputSettingsOnFileThread,
 688                                      audio_manager_));
 689 }
 690
 691 SpeechRecognitionManagerImpl::Session::Session()
 692   : id(kSessionIDInvalid),
 693     abort_requested(false),
 694     listener_is_active(true) {
 695 }
 696
 697 SpeechRecognitionManagerImpl::Session::~Session() {
 698 }
 699
 700 }  // namespace content