Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(73)

Side by Side Diff: content/browser/speech/speech_recognizer.h

Issue 8137005: Applying changes to the existing speech input code to support the extension API. (Closed) Base URL: svn://svn.chromium.org/chrome/trunk/src
Patch Set: review and unit test fixes. Created 9 years, 2 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch | Annotate | Revision Log
OLDNEW
1 // Copyright (c) 2011 The Chromium Authors. All rights reserved. 1 // Copyright (c) 2011 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be 2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file. 3 // found in the LICENSE file.
4 4
5 #ifndef CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_H_ 5 #ifndef CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_H_
6 #define CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_H_ 6 #define CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_H_
7 7
8 #include <list> 8 #include <list>
9 #include <string> 9 #include <string>
10 #include <utility> 10 #include <utility>
(...skipping 12 matching lines...) Expand all
23 23
24 namespace speech_input { 24 namespace speech_input {
25 25
26 // Records audio, sends recorded audio to server and translates server response 26 // Records audio, sends recorded audio to server and translates server response
27 // to recognition result. 27 // to recognition result.
28 class CONTENT_EXPORT SpeechRecognizer 28 class CONTENT_EXPORT SpeechRecognizer
29 : public base::RefCountedThreadSafe<SpeechRecognizer>, 29 : public base::RefCountedThreadSafe<SpeechRecognizer>,
30 public media::AudioInputController::EventHandler, 30 public media::AudioInputController::EventHandler,
31 public SpeechRecognitionRequestDelegate { 31 public SpeechRecognitionRequestDelegate {
32 public: 32 public:
33 enum ErrorCode {
34 RECOGNIZER_NO_ERROR,
35 RECOGNIZER_ERROR_CAPTURE,
36 RECOGNIZER_ERROR_NO_SPEECH,
37 RECOGNIZER_ERROR_NO_RESULTS,
38 RECOGNIZER_ERROR_NETWORK,
39 };
40
41 // Implemented by the caller to receive recognition events. 33 // Implemented by the caller to receive recognition events.
42 class CONTENT_EXPORT Delegate { 34 class CONTENT_EXPORT Delegate {
43 public: 35 public:
44 virtual void SetRecognitionResult( 36 virtual void SetRecognitionResult(
45 int caller_id, 37 int caller_id,
46 bool error, 38 bool error,
47 const SpeechInputResultArray& result) = 0; 39 const SpeechInputResult& result) = 0;
48 40
49 // Invoked when the first audio packet was received from the audio capture 41 // Invoked when the first audio packet was received from the audio capture
50 // device. 42 // device.
51 virtual void DidStartReceivingAudio(int caller_id) = 0; 43 virtual void DidStartReceivingAudio(int caller_id) = 0;
52 44
53 // Invoked when audio recording stops, either due to the end pointer 45 // Invoked when audio recording stops, either due to the end pointer
54 // detecting silence in user input or if |StopRecording| was called. The 46 // detecting silence in user input or if |StopRecording| was called. The
55 // delegate has to wait until |DidCompleteRecognition| is invoked before 47 // delegate has to wait until |DidCompleteRecognition| is invoked before
56 // destroying the |SpeechRecognizer| object. 48 // destroying the |SpeechRecognizer| object.
57 virtual void DidCompleteRecording(int caller_id) = 0; 49 virtual void DidCompleteRecording(int caller_id) = 0;
58 50
59 // This is guaranteed to be the last method invoked in the recognition 51 // This is guaranteed to be the last method invoked in the recognition
60 // sequence and the |SpeechRecognizer| object can be freed up if necessary. 52 // sequence and the |SpeechRecognizer| object can be freed up if necessary.
61 virtual void DidCompleteRecognition(int caller_id) = 0; 53 virtual void DidCompleteRecognition(int caller_id) = 0;
62 54
55 // Informs that the end pointer has started detecting input speech.
Satish 2011/10/06 09:09:06 "input speech" -> "speech"
Leandro Graciá Gil 2011/10/06 18:26:25 Done.
56 virtual void DidStartReceivingSpeech(int caller_id) = 0;
57
58 // Informs that the end pointer has stopped detecting input speech.
Satish 2011/10/06 09:09:06 ditto
Leandro Graciá Gil 2011/10/06 18:26:25 Done.
59 virtual void DidStopReceivingSpeech(int caller_id) = 0;
60
63 // Invoked if there was an error while recording or recognizing audio. The 61 // Invoked if there was an error while recording or recognizing audio. The
64 // session has already been cancelled when this call is made and the DidXxxx 62 // session has already been cancelled when this call is made and the DidXxxx
65 // callbacks will not be issued. It is safe to destroy/release the 63 // callbacks will not be issued. It is safe to destroy/release the
66 // |SpeechRecognizer| object while processing this call. 64 // |SpeechRecognizer| object while processing this call.
67 virtual void OnRecognizerError(int caller_id, 65 virtual void OnRecognizerError(int caller_id,
68 SpeechRecognizer::ErrorCode error) = 0; 66 SpeechInputError error) = 0;
69 67
70 // At the start of recognition, a short amount of audio is recorded to 68 // At the start of recognition, a short amount of audio is recorded to
71 // estimate the environment/background noise and this callback is issued 69 // estimate the environment/background noise and this callback is issued
72 // after that is complete. Typically the delegate brings up any speech 70 // after that is complete. Typically the delegate brings up any speech
73 // recognition UI once this callback is received. 71 // recognition UI once this callback is received.
74 virtual void DidCompleteEnvironmentEstimation(int caller_id) = 0; 72 virtual void DidCompleteEnvironmentEstimation(int caller_id) = 0;
75 73
76 // Informs of a change in the captured audio level, useful if displaying 74 // Informs of a change in the captured audio level, useful if displaying
77 // a microphone volume indicator while recording. 75 // a microphone volume indicator while recording.
78 // The value of |volume| and |noise_volume| is in the [0.0, 1.0] range. 76 // The value of |volume| and |noise_volume| is in the [0.0, 1.0] range.
79 virtual void SetInputVolume(int caller_id, float volume, 77 virtual void SetInputVolume(int caller_id, float volume,
80 float noise_volume) = 0; 78 float noise_volume) = 0;
81 79
82 protected: 80 protected:
83 virtual ~Delegate() {} 81 virtual ~Delegate() {}
84 }; 82 };
85 83
86 SpeechRecognizer(Delegate* delegate, 84 SpeechRecognizer(Delegate* delegate,
87 int caller_id, 85 int caller_id,
88 const std::string& language, 86 const std::string& language,
89 const std::string& grammar, 87 const std::string& grammar,
90 net::URLRequestContextGetter* context_getter, 88 net::URLRequestContextGetter* context_getter,
91 bool censor_results, 89 bool censor_results,
92 const std::string& hardware_info, 90 const std::string& hardware_info,
93 const std::string& origin_url); 91 const std::string& origin_url);
92
94 virtual ~SpeechRecognizer(); 93 virtual ~SpeechRecognizer();
95 94
96 // Starts audio recording and does recognition after recording ends. The same 95 // Starts audio recording and does recognition after recording ends. The same
97 // SpeechRecognizer instance can be used multiple times for speech recognition 96 // SpeechRecognizer instance can be used multiple times for speech recognition
98 // though each recognition request can be made only after the previous one 97 // though each recognition request can be made only after the previous one
99 // completes (i.e. after receiving Delegate::DidCompleteRecognition). 98 // completes (i.e. after receiving Delegate::DidCompleteRecognition).
100 bool StartRecording(); 99 bool StartRecording();
101 100
102 // Stops recording audio and starts recognition. 101 // Stops recording audio and starts recognition.
103 void StopRecording(); 102 void StopRecording();
104 103
105 // Stops recording audio and cancels recognition. Any audio recorded so far 104 // Stops recording audio and cancels recognition. Any audio recorded so far
106 // gets discarded. 105 // gets discarded.
107 void CancelRecognition(); 106 void CancelRecognition();
108 107
109 // AudioInputController::EventHandler methods. 108 // AudioInputController::EventHandler methods.
110 virtual void OnCreated(media::AudioInputController* controller) { } 109 virtual void OnCreated(media::AudioInputController* controller) { }
111 virtual void OnRecording(media::AudioInputController* controller) { } 110 virtual void OnRecording(media::AudioInputController* controller) { }
112 virtual void OnError(media::AudioInputController* controller, int error_code); 111 virtual void OnError(media::AudioInputController* controller, int error_code);
113 virtual void OnData(media::AudioInputController* controller, 112 virtual void OnData(media::AudioInputController* controller,
114 const uint8* data, 113 const uint8* data,
115 uint32 size); 114 uint32 size);
116 115
117 // SpeechRecognitionRequest::Delegate methods. 116 // SpeechRecognitionRequest::Delegate methods.
118 virtual void SetRecognitionResult(bool error, 117 virtual void SetRecognitionResult(bool error,
119 const SpeechInputResultArray& result); 118 const SpeechInputResult& result);
Satish 2011/10/06 09:09:06 could align with the previous param?
Leandro Graciá Gil 2011/10/06 18:26:25 Done.
120 119
121 static const int kAudioSampleRate; 120 static const int kAudioSampleRate;
122 static const int kAudioPacketIntervalMs; // Duration of each audio packet. 121 static const int kAudioPacketIntervalMs; // Duration of each audio packet.
123 static const ChannelLayout kChannelLayout; 122 static const ChannelLayout kChannelLayout;
124 static const int kNumBitsPerAudioSample; 123 static const int kNumBitsPerAudioSample;
125 static const int kNoSpeechTimeoutSec; 124 static const int kNoSpeechTimeoutSec;
126 static const int kEndpointerEstimationTimeMs; 125 static const int kEndpointerEstimationTimeMs;
127 126
128 private: 127 private:
129 void InformErrorAndCancelRecognition(ErrorCode error); 128 void InformErrorAndCancelRecognition(SpeechInputError error);
130 void SendRecordedAudioToServer(); 129 void SendRecordedAudioToServer();
131 130
132 void HandleOnError(int error_code); // Handles OnError in the IO thread. 131 void HandleOnError(int error_code); // Handles OnError in the IO thread.
133 132
134 // Handles OnData in the IO thread. Takes ownership of |data|. 133 // Handles OnData in the IO thread. Takes ownership of |data|.
135 void HandleOnData(std::string* data); 134 void HandleOnData(std::string* data);
136 135
137 Delegate* delegate_; 136 Delegate* delegate_;
138 int caller_id_; 137 int caller_id_;
139 std::string language_; 138 std::string language_;
(...skipping 16 matching lines...) Expand all
156 155
157 // This typedef is to workaround the issue with certain versions of 156 // This typedef is to workaround the issue with certain versions of
158 // Visual Studio where it gets confused between multiple Delegate 157 // Visual Studio where it gets confused between multiple Delegate
159 // classes and gives a C2500 error. (I saw this error on the try bots - 158 // classes and gives a C2500 error. (I saw this error on the try bots -
160 // the workaround was not needed for my machine). 159 // the workaround was not needed for my machine).
161 typedef SpeechRecognizer::Delegate SpeechRecognizerDelegate; 160 typedef SpeechRecognizer::Delegate SpeechRecognizerDelegate;
162 161
163 } // namespace speech_input 162 } // namespace speech_input
164 163
165 #endif // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_H_ 164 #endif // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_H_
OLDNEW

Powered by Google App Engine
This is Rietveld 408576698