Applying changes to the existing speech input code to support the extension API.

content/browser/speech/speech_recognizer.h

 // Copyright (c) 2011 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_H_
 #define CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_H_
 
 #include <list>
 #include <string>
 #include <utility>
@@ skipping 12 matching lines @@
 
 namespace speech_input {
 
 // Records audio, sends recorded audio to server and translates server response
 // to recognition result.
 class CONTENT_EXPORT SpeechRecognizer
     : public base::RefCountedThreadSafe<SpeechRecognizer>,
       public media::AudioInputController::EventHandler,
       public SpeechRecognitionRequestDelegate {
  public:
   enum ErrorCode {

[Satish, 2011/10/04 20:36:33]

Can this enum be removed and we just use SpeechInputResultStatus instead?

     RECOGNIZER_NO_ERROR,
     RECOGNIZER_ERROR_CAPTURE,
     RECOGNIZER_ERROR_NO_SPEECH,
     RECOGNIZER_ERROR_NO_RESULTS,
     RECOGNIZER_ERROR_NETWORK,
+    RECOGNIZER_ERROR_INVALID_PARAMS,
+    RECOGNIZER_ERROR_INTERNAL,

[Satish, 2011/10/04 20:36:33]

since this is related to the protocol/parse error, i'd suggest naming it as
protocol_error, or use ERROR_NETWORK instead

[Leandro Graciá Gil, 2011/10/05 22:09:00]

I'll go for network error as it's an un expected value for the server response.

On 2011/10/04 20:36:33, Satish wrote:

   };
 
   // Implemented by the caller to receive recognition events.
   class CONTENT_EXPORT Delegate {
    public:
     virtual void SetRecognitionResult(
         int caller_id,
         bool error,
-        const SpeechInputResultArray& result) = 0;
+        const SpeechInputResult& result) = 0;
 
     // Invoked when the first audio packet was received from the audio capture
     // device.
     virtual void DidStartReceivingAudio(int caller_id) = 0;
 
     // Invoked when audio recording stops, either due to the end pointer
     // detecting silence in user input or if |StopRecording| was called. The
     // delegate has to wait until |DidCompleteRecognition| is invoked before
     // destroying the |SpeechRecognizer| object.
     virtual void DidCompleteRecording(int caller_id) = 0;
 
     // This is guaranteed to be the last method invoked in the recognition
     // sequence and the |SpeechRecognizer| object can be freed up if necessary.
     virtual void DidCompleteRecognition(int caller_id) = 0;
 
+    // Informs that the end pointer has started detecting input speech.
+    virtual void DidSpeechInputStart(int caller_id) = 0;

[Satish, 2011/10/04 20:36:33]

this would be clearer if renamed as DidStartReceivingSpeech and similarly the
one below

[Leandro Graciá Gil, 2011/10/05 22:09:00]

Done.

+
+    // Informs that the end pointer has stopped detecting input speech.
+    virtual void DidSpeechInputStop(int caller_id) = 0;
+
+    // Invoked when the recording is actually started.
+    virtual void OnRecording(int caller_id) = 0;

[Satish, 2011/10/04 20:36:33]

any reason why we need this when we have DidStartReceivingAudio? Is it really
important to inform about this and not wait for the first packet of audio (which
could be a more accurate representation of audio recording start) ?

[Leandro Graciá Gil, 2011/10/05 22:09:00]

Fixed. Now using only DidStartReceivingAudio after checking that it's properly
called even when the mic is muted.

On 2011/10/04 20:36:33, Satish wrote:

+
     // Invoked if there was an error while recording or recognizing audio. The
     // session has already been cancelled when this call is made and the DidXxxx
     // callbacks will not be issued. It is safe to destroy/release the
     // |SpeechRecognizer| object while processing this call.
     virtual void OnRecognizerError(int caller_id,
                                    SpeechRecognizer::ErrorCode error) = 0;
 
     // At the start of recognition, a short amount of audio is recorded to
     // estimate the environment/background noise and this callback is issued
     // after that is complete. Typically the delegate brings up any speech
@@ skipping 11 matching lines @@
   };
 
   SpeechRecognizer(Delegate* delegate,
                    int caller_id,
                    const std::string& language,
                    const std::string& grammar,
                    net::URLRequestContextGetter* context_getter,
                    bool censor_results,
                    const std::string& hardware_info,
                    const std::string& origin_url);
+
   virtual ~SpeechRecognizer();
 
   // Starts audio recording and does recognition after recording ends. The same
   // SpeechRecognizer instance can be used multiple times for speech recognition
   // though each recognition request can be made only after the previous one
   // completes (i.e. after receiving Delegate::DidCompleteRecognition).
   bool StartRecording();
 
   // Stops recording audio and starts recognition.
   void StopRecording();
 
   // Stops recording audio and cancels recognition. Any audio recorded so far
   // gets discarded.
   void CancelRecognition();
 
   // AudioInputController::EventHandler methods.
   virtual void OnCreated(media::AudioInputController* controller) { }
-  virtual void OnRecording(media::AudioInputController* controller) { }
+  virtual void OnRecording(media::AudioInputController* controller);
   virtual void OnError(media::AudioInputController* controller, int error_code);
   virtual void OnData(media::AudioInputController* controller,
                       const uint8* data,
                       uint32 size);
 
   // SpeechRecognitionRequest::Delegate methods.
   virtual void SetRecognitionResult(bool error,
-                                    const SpeechInputResultArray& result);
+      const SpeechInputResult& result);
 
   static const int kAudioSampleRate;
   static const int kAudioPacketIntervalMs;  // Duration of each audio packet.
   static const ChannelLayout kChannelLayout;
   static const int kNumBitsPerAudioSample;
   static const int kNoSpeechTimeoutSec;
   static const int kEndpointerEstimationTimeMs;
 
  private:
   void InformErrorAndCancelRecognition(ErrorCode error);
   void SendRecordedAudioToServer();
 
+  void HandleOnRecording();  // Handles OnRecording in the IO thread.
   void HandleOnError(int error_code);  // Handles OnError in the IO thread.
 
   // Handles OnData in the IO thread. Takes ownership of |data|.
   void HandleOnData(std::string* data);
 
   Delegate* delegate_;
   int caller_id_;
   std::string language_;
   std::string grammar_;
   bool censor_results_;
@@ skipping 14 matching lines @@
 
 // This typedef is to workaround the issue with certain versions of
 // Visual Studio where it gets confused between multiple Delegate
 // classes and gives a C2500 error. (I saw this error on the try bots -
 // the workaround was not needed for my machine).
 typedef SpeechRecognizer::Delegate SpeechRecognizerDelegate;
 
 }  // namespace speech_input
 
 #endif  // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_H_