Speech refactoring: Reimplemented speech_recognizer as a FSM. (CL1.5)

content/browser/speech/speech_recognizer_impl.h

 // Copyright (c) 2012 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/ref_counted.h"
 #include "base/memory/scoped_ptr.h"
 #include "content/browser/speech/endpointer/endpointer.h"
 #include "content/browser/speech/speech_recognition_engine.h"
 #include "content/public/browser/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 "net/url_request/url_request_context_getter.h"
 
 namespace content {
 class SpeechRecognitionEventListener;
-struct SpeechRecognitionResult;
-}
-
-namespace media {
-class AudioInputController;
 }
 
 namespace speech {
+// TODO(primiano) Next CL: Remove the Impl suffix and the exported
+// /content/public/browser/speech_recognizer.h interface since this class should
+// not be visible outside (currently we need it for speech input extension API).
 
-// Records audio, sends recorded audio to server and translates server response
-// to recognition result.
+// Handles speech recognition for a session (identified by |caller_id|), taking
+// care of audio capture, silence detection/endpointer and interaction with the
+// SpeechRecognitionEngine.
 class CONTENT_EXPORT SpeechRecognizerImpl
     : public NON_EXPORTED_BASE(content::SpeechRecognizer),
       public media::AudioInputController::EventHandler,
       public NON_EXPORTED_BASE(SpeechRecognitionEngineDelegate) {
  public:
   static const int kAudioSampleRate;
   static const ChannelLayout kChannelLayout;
   static const int kNumBitsPerAudioSample;
   static const int kNoSpeechTimeoutMs;
   static const int kEndpointerEstimationTimeMs;
 
   SpeechRecognizerImpl(
-    content::SpeechRecognitionEventListener* listener,
-    int caller_id,
-    const std::string& language,
-    const std::string& grammar,
-    net::URLRequestContextGetter* context_getter,
-    bool filter_profanities,
-    const std::string& hardware_info,
-    const std::string& origin_url);
+      content::SpeechRecognitionEventListener* listener,
+      int caller_id,
+      SpeechRecognitionEngine* engine);
   virtual ~SpeechRecognizerImpl();
 
   // content::SpeechRecognizer methods.
   virtual void StartRecognition() OVERRIDE;
   virtual void AbortRecognition() OVERRIDE;
   virtual void StopAudioCapture() OVERRIDE;
   virtual bool IsActive() const OVERRIDE;
   virtual bool IsCapturingAudio() const OVERRIDE;
   const SpeechRecognitionEngine& recognition_engine() const;
 
+ private:
+  friend class SpeechRecognizerImplTest;
+
+  enum FSMState {
+    STATE_IDLE = 0,
+    STATE_STARTING,
+    STATE_ESTIMATING_ENVIRONMENT,
+    STATE_WAITING_FOR_SPEECH,
+    STATE_RECOGNIZING,
+    STATE_WAITING_FINAL_RESULT,
+    STATE_MAX = STATE_WAITING_FINAL_RESULT
+  };
+
+  enum FSMEvent {
+    EVENT_ABORT = 0,
+    EVENT_START,
+    EVENT_STOP_CAPTURE,
+    EVENT_AUDIO_DATA,
+    EVENT_ENGINE_RESULT,
+    EVENT_ENGINE_ERROR,
+    EVENT_AUDIO_ERROR,
+    EVENT_MAX = EVENT_AUDIO_ERROR
+  };
+
+  struct FSMEventArgs {
+    FSMEvent event;
+    int audio_error_code;
+    scoped_refptr<AudioChunk> audio_data;
+    content::SpeechRecognitionResult engine_result;
+    content::SpeechRecognitionError engine_error;
+    FSMEventArgs();
+    ~FSMEventArgs();
+  };
+
+  typedef base::Callback<FSMState(const FSMEventArgs&)> TransitionFunction;
+
   // AudioInputController::EventHandler methods.
   virtual void OnCreated(media::AudioInputController* controller) OVERRIDE {}
   virtual void OnRecording(media::AudioInputController* controller) OVERRIDE {}
   virtual void OnError(media::AudioInputController* controller,
                        int error_code) OVERRIDE;
   virtual void OnData(media::AudioInputController* controller,
-                      const uint8* data,
-                      uint32 size) OVERRIDE;
+                      const uint8* data, uint32 size) OVERRIDE;
+  // Callback called on IO thread by audio_controller->Close().

[hans, 2012/04/02 16:05:59]

nit: maybe put a blank line before this, since it's not part of the EventHandler
methods.

[Primiano Tucci (use gerrit), 2012/04/03 10:16:39]

Done.

+  void OnAudioClosed(media::AudioInputController*);
 
   // SpeechRecognitionEngineDelegate methods.
   virtual void OnSpeechRecognitionEngineResult(
       const content::SpeechRecognitionResult& result) OVERRIDE;
   virtual void OnSpeechRecognitionEngineError(
       const content::SpeechRecognitionError& error) OVERRIDE;
 
- private:
-  friend class SpeechRecognizerImplTest;
-
-  void InformErrorAndAbortRecognition(
-      content::SpeechRecognitionErrorCode error);
-  void SendRecordedAudioToServer();
-
-  void HandleOnError(int error_code);  // Handles OnError in the IO thread.
-
-  // Handles OnData in the IO thread.
-  void HandleOnData(scoped_refptr<AudioChunk> raw_audio);
-
-  void OnAudioClosed(media::AudioInputController*);
-
-  // Helper method which closes the audio controller and frees it asynchronously
-  // without blocking the IO thread.
+  void InitializeFSM();
+  void DispatchEvent(FSMEvent, FSMEventArgs);
+  void ProcessAudioPipeline(const AudioChunk& raw_audio);
+  FSMState StartRecording(const FSMEventArgs& event_args);
+  FSMState StartRecognitionEngine(const FSMEventArgs& event_args);
+  FSMState WaitEnvironmentEstimationCompletion(const FSMEventArgs& event_args);
+  FSMState DetectUserSpeechOrTimeout(const FSMEventArgs& event_args);
+  FSMState StopCaptureAndWaitResult(const FSMEventArgs& event_args);
+  FSMState ProcessIntermediateResult(const FSMEventArgs& event_args);
+  FSMState ProcessFinalResult(const FSMEventArgs& event_args);
+  FSMState Abort(const FSMEventArgs& event_args);
+  FSMState AbortWithError(const content::SpeechRecognitionError* error);
+  FSMState AbortWithError(const content::SpeechRecognitionError& error);
+  FSMState DetectEndOfSpeech(const FSMEventArgs& event_args);
+  FSMState DoNothing(const FSMEventArgs& event_args) const;
+  int GetElapsedTimeMs() const;
+  void UpdateSignalAndNoiseLevels(const float& rms, bool clip_detected);
   void CloseAudioControllerAsynchronously();
-
   void SetAudioManagerForTesting(AudioManager* audio_manager);
 
   content::SpeechRecognitionEventListener* listener_;
   AudioManager* testing_audio_manager_;
   scoped_ptr<SpeechRecognitionEngine> recognition_engine_;
   Endpointer endpointer_;
   scoped_refptr<media::AudioInputController> audio_controller_;
-  scoped_refptr<net::URLRequestContextGetter> context_getter_;
   int caller_id_;
-  std::string language_;
-  std::string grammar_;
-  bool filter_profanities_;
-  std::string hardware_info_;
-  std::string origin_url_;
   int num_samples_recorded_;
   float audio_level_;
+  bool in_event_dispatching_;
+  FSMState state_;
+  TransitionFunction fsm[STATE_MAX + 1][EVENT_MAX + 1];
+  const TransitionFunction kUnfeasibleTransition;
 
   DISALLOW_COPY_AND_ASSIGN(SpeechRecognizerImpl);
 };
 
 }  // namespace speech
 
 #endif  // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_