Binding media stream audio track to speech recognition [browser]

content/browser/speech/speech_recognizer_impl.h

 // Copyright (c) 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #ifndef CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_
 #define CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_
 
 #include "base/basictypes.h"
 #include "base/memory/scoped_ptr.h"
+#include "base/memory/shared_memory.h"
+#include "base/process/process_handle.h"
+#include "base/sync_socket.h"
+#include "base/threading/thread.h"
 #include "content/browser/speech/endpointer/endpointer.h"
 #include "content/browser/speech/speech_recognition_engine.h"
 #include "content/browser/speech/speech_recognizer.h"
 #include "content/public/common/speech_recognition_error.h"
 #include "content/public/common/speech_recognition_result.h"
 #include "media/audio/audio_input_controller.h"
 #include "media/audio/audio_logging.h"
 #include "net/url_request/url_request_context_getter.h"
 
 namespace media {
 class AudioBus;
 class AudioManager;
+class AudioParameters;
 }
 
 namespace content {
 
 class SpeechRecognitionEventListener;
 
 // Handles speech recognition for a session (identified by |session_id|), taking
 // care of audio capture, silence detection/endpointer and interaction with the
 // SpeechRecognitionEngine.
 class CONTENT_EXPORT SpeechRecognizerImpl
     : public SpeechRecognizer,
       public media::AudioInputController::EventHandler,
       public NON_EXPORTED_BASE(SpeechRecognitionEngineDelegate) {
  public:
   static const int kAudioSampleRate;
   static const media::ChannelLayout kChannelLayout;
   static const int kNumBitsPerAudioSample;
   static const int kNoSpeechTimeoutMs;
   static const int kEndpointerEstimationTimeMs;
 
   static void SetAudioManagerForTesting(media::AudioManager* audio_manager);
-
-  SpeechRecognizerImpl(SpeechRecognitionEventListener* listener,
-                       int session_id,
-                       bool continuous,
-                       bool provisional_results,
+  // TODO(burnik): abstract out the process_handle. Pass differently.
+  SpeechRecognizerImpl(SpeechRecognitionEventListener* listener, int session_id,
+                       bool continuous, bool provisional_results,
+                       bool using_audio_track,
+                       base::ProcessHandle process_handle,
                        SpeechRecognitionEngine* engine);
 
   virtual void StartRecognition(const std::string& device_id) OVERRIDE;
   virtual void AbortRecognition() OVERRIDE;
   virtual void StopAudioCapture() OVERRIDE;
   virtual bool IsActive() const OVERRIDE;
   virtual bool IsCapturingAudio() const OVERRIDE;
   const SpeechRecognitionEngine& recognition_engine() const;
 
+  // Whether to use an audio track from the renderer as the input source.
+  void set_using_audio_track(bool using_audio_track) {

[tommi (sloooow) - chröme, 2014/10/02 11:23:19]

thread check?
(and move to cc file)

[burnik, 2014/10/02 16:04:12]

Done.

+    using_audio_track_ = using_audio_track;
+  }
+
  private:
   friend class SpeechRecognizerTest;
 
   enum FSMState {
     STATE_IDLE = 0,
     STATE_STARTING,
     STATE_ESTIMATING_ENVIRONMENT,
     STATE_WAITING_FOR_SPEECH,
     STATE_RECOGNIZING,
     STATE_WAITING_FINAL_RESULT,
@@ skipping 42 matching lines @@
   FSMState StopCaptureAndWaitForResult(const FSMEventArgs& event_args);
   FSMState ProcessIntermediateResult(const FSMEventArgs& event_args);
   FSMState ProcessFinalResult(const FSMEventArgs& event_args);
   FSMState AbortSilently(const FSMEventArgs& event_args);
   FSMState AbortWithError(const FSMEventArgs& event_args);
   FSMState Abort(const SpeechRecognitionError& error);
   FSMState DetectEndOfSpeech(const FSMEventArgs& event_args);
   FSMState DoNothing(const FSMEventArgs& event_args) const;
   FSMState NotFeasible(const FSMEventArgs& event_args);
 
+  // Called from |StartRecording()| when using audio track from renderer.
+  bool StartRecordingUsingAudioTrack(const media::AudioParameters& params);
+
   // Returns the time span of captured audio samples since the start of capture.
   int GetElapsedTimeMs() const;
 
   // Calculates the input volume to be displayed in the UI, triggering the
   // OnAudioLevelsChange event accordingly.
   void UpdateSignalAndNoiseLevels(const float& rms, bool clip_detected);
 
   void CloseAudioControllerAsynchronously();
 
   // Callback called on IO thread by audio_controller->Close().
@@ skipping 18 matching lines @@
   static media::AudioManager* audio_manager_for_tests_;
 
   scoped_ptr<SpeechRecognitionEngine> recognition_engine_;
   Endpointer endpointer_;
   scoped_refptr<media::AudioInputController> audio_controller_;
   scoped_ptr<media::AudioLog> audio_log_;
   int num_samples_recorded_;
   float audio_level_;
   bool is_dispatching_event_;
   bool provisional_results_;
+  bool using_audio_track_;
+  base::ProcessHandle peer_process_handle_;
   FSMState state_;
   std::string device_id_;
+  scoped_ptr<base::CancelableSyncSocket> audio_sync_socket_;
+  scoped_ptr<base::CancelableSyncSocket> foreign_sync_socket_;
+
+  // Runs on separate |audio_sync_thread_| when receiving audio from renderer.
+  virtual void OnAudioTrackData();
 
   class OnDataConverter;
 
   // Converts data between native input format and a WebSpeech specific
   // output format.
   scoped_ptr<SpeechRecognizerImpl::OnDataConverter> audio_converter_;
 
+  // Shared memory for passing audio data from renderer to the browser.
+  scoped_ptr<base::SharedMemory> shared_memory_;
+
+  // Audio bus wrapping the shared memory from the renderer.
+  scoped_ptr<media::AudioBus> audio_track_bus_;
+
+  // Used for synchronization when consuming audio on the |audio_track_bus_|.
+  uint32* buffer_index_;
+
+  // Thread for receiving audio chunks from the renderer.
+  scoped_ptr<base::Thread> audio_sync_thread_;
+
+  // Used to check if |OnAudioTrackData()| is executed on |audio_sync_thread_|.
+  base::ThreadChecker audio_sync_thread_checker_;
+
   DISALLOW_COPY_AND_ASSIGN(SpeechRecognizerImpl);
 };
 
 }  // namespace content
 
 #endif  // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_