OLD | NEW |
1 // Copyright (c) 2012 The Chromium Authors. All rights reserved. | 1 // Copyright (c) 2012 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_IMPL_H_ | 5 #ifndef CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_ |
6 #define CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_ | 6 #define CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_ |
7 | 7 |
8 #include <list> | 8 #include <list> |
9 #include <utility> | 9 #include <utility> |
10 | 10 |
11 #include "base/compiler_specific.h" | 11 #include "base/compiler_specific.h" |
12 #include "base/memory/scoped_ptr.h" | 12 #include "base/memory/scoped_ptr.h" |
13 #include "content/browser/speech/audio_encoder.h" | 13 #include "content/browser/speech/audio_encoder.h" |
14 #include "content/browser/speech/endpointer/endpointer.h" | 14 #include "content/browser/speech/endpointer/endpointer.h" |
15 #include "content/browser/speech/speech_recognition_request.h" | 15 #include "content/browser/speech/speech_recognition_request.h" |
16 #include "content/public/browser/speech_recognizer.h" | 16 #include "content/public/browser/speech_recognizer.h" |
17 #include "content/public/common/speech_recognition_result.h" | 17 #include "content/public/common/speech_recognition_result.h" |
18 #include "media/audio/audio_input_controller.h" | 18 #include "media/audio/audio_input_controller.h" |
19 | 19 |
20 class AudioManager; | 20 class AudioManager; |
21 | 21 |
| 22 namespace content { |
| 23 class SpeechRecognitionEventListener; |
| 24 } |
| 25 |
22 namespace speech { | 26 namespace speech { |
23 | 27 |
24 // Records audio, sends recorded audio to server and translates server response | 28 // Records audio, sends recorded audio to server and translates server response |
25 // to recognition result. | 29 // to recognition result. |
26 class CONTENT_EXPORT SpeechRecognizerImpl | 30 class CONTENT_EXPORT SpeechRecognizerImpl |
27 : NON_EXPORTED_BASE(public content::SpeechRecognizer), | 31 : NON_EXPORTED_BASE(public content::SpeechRecognizer), |
28 public media::AudioInputController::EventHandler, | 32 public media::AudioInputController::EventHandler, |
29 public SpeechRecognitionRequestDelegate { | 33 public SpeechRecognitionRequestDelegate { |
30 public: | 34 public: |
31 SpeechRecognizerImpl(content::SpeechRecognizerDelegate* delegate, | 35 static const int kAudioSampleRate; |
| 36 static const int kAudioPacketIntervalMs; // Duration of each audio packet. |
| 37 static const ChannelLayout kChannelLayout; |
| 38 static const int kNumBitsPerAudioSample; |
| 39 static const int kNoSpeechTimeoutSec; |
| 40 static const int kEndpointerEstimationTimeMs; |
| 41 |
| 42 SpeechRecognizerImpl(content::SpeechRecognitionEventListener* listener, |
32 int caller_id, | 43 int caller_id, |
33 const std::string& language, | 44 const std::string& language, |
34 const std::string& grammar, | 45 const std::string& grammar, |
35 net::URLRequestContextGetter* context_getter, | 46 net::URLRequestContextGetter* context_getter, |
36 bool filter_profanities, | 47 bool filter_profanities, |
37 const std::string& hardware_info, | 48 const std::string& hardware_info, |
38 const std::string& origin_url); | 49 const std::string& origin_url); |
39 | 50 |
40 virtual ~SpeechRecognizerImpl(); | 51 virtual ~SpeechRecognizerImpl(); |
41 | 52 |
42 // SpeechRecognizer implementation: | 53 // content::SpeechRecognizer methods. |
43 virtual bool StartRecording() OVERRIDE; | 54 virtual bool StartRecognition() OVERRIDE; |
44 virtual void CancelRecognition() OVERRIDE; | 55 virtual void AbortRecognition() OVERRIDE; |
45 | 56 virtual void StopAudioCapture() OVERRIDE; |
46 // Stops recording audio and starts recognition. | 57 virtual bool IsActive() const OVERRIDE; |
47 void StopRecording(); | 58 virtual bool IsCapturingAudio() const OVERRIDE; |
48 | 59 |
49 // AudioInputController::EventHandler methods. | 60 // AudioInputController::EventHandler methods. |
50 virtual void OnCreated(media::AudioInputController* controller) OVERRIDE {} | 61 virtual void OnCreated(media::AudioInputController* controller) OVERRIDE {} |
51 virtual void OnRecording(media::AudioInputController* controller) OVERRIDE {} | 62 virtual void OnRecording(media::AudioInputController* controller) OVERRIDE {} |
52 virtual void OnError(media::AudioInputController* controller, | 63 virtual void OnError(media::AudioInputController* controller, |
53 int error_code) OVERRIDE; | 64 int error_code) OVERRIDE; |
54 virtual void OnData(media::AudioInputController* controller, | 65 virtual void OnData(media::AudioInputController* controller, |
55 const uint8* data, | 66 const uint8* data, |
56 uint32 size) OVERRIDE; | 67 uint32 size) OVERRIDE; |
57 | 68 |
58 // SpeechRecognitionRequest::Delegate methods. | 69 // SpeechRecognitionRequest::Delegate methods. |
59 virtual void SetRecognitionResult( | 70 virtual void SetRecognitionResult( |
60 const content::SpeechRecognitionResult& result) OVERRIDE; | 71 const content::SpeechRecognitionResult& result) OVERRIDE; |
61 | 72 |
62 static const int kAudioSampleRate; | 73 private: |
63 static const int kAudioPacketIntervalMs; // Duration of each audio packet. | 74 friend class SpeechRecognizerImplTest; |
64 static const ChannelLayout kChannelLayout; | |
65 static const int kNumBitsPerAudioSample; | |
66 static const int kNoSpeechTimeoutSec; | |
67 static const int kEndpointerEstimationTimeMs; | |
68 | 75 |
69 private: | 76 void InformErrorAndAbortRecognition( |
70 friend class SpeechRecognizerTest; | |
71 | |
72 void InformErrorAndCancelRecognition( | |
73 content::SpeechRecognitionErrorCode error); | 77 content::SpeechRecognitionErrorCode error); |
74 void SendRecordedAudioToServer(); | 78 void SendRecordedAudioToServer(); |
75 | 79 |
76 void HandleOnError(int error_code); // Handles OnError in the IO thread. | 80 void HandleOnError(int error_code); // Handles OnError in the IO thread. |
77 | 81 |
78 // Handles OnData in the IO thread. Takes ownership of |raw_audio|. | 82 // Handles OnData in the IO thread. Takes ownership of |raw_audio|. |
79 void HandleOnData(AudioChunk* raw_audio); | 83 void HandleOnData(AudioChunk* raw_audio); |
80 | 84 |
81 // Helper method which closes the audio controller and blocks until done. | 85 // Helper method which closes the audio controller and blocks until done. |
82 void CloseAudioControllerSynchronously(); | 86 void CloseAudioControllerSynchronously(); |
83 | 87 |
84 void SetAudioManagerForTesting(AudioManager* audio_manager); | 88 void SetAudioManagerForTesting(AudioManager* audio_manager); |
85 | 89 |
86 content::SpeechRecognizerDelegate* delegate_; | 90 content::SpeechRecognitionEventListener* listener_; |
87 int caller_id_; | 91 int caller_id_; |
88 std::string language_; | 92 std::string language_; |
89 std::string grammar_; | 93 std::string grammar_; |
90 bool filter_profanities_; | 94 bool filter_profanities_; |
91 std::string hardware_info_; | 95 std::string hardware_info_; |
92 std::string origin_url_; | 96 std::string origin_url_; |
93 | 97 |
94 scoped_ptr<SpeechRecognitionRequest> request_; | 98 scoped_ptr<SpeechRecognitionRequest> request_; |
95 scoped_refptr<media::AudioInputController> audio_controller_; | 99 scoped_refptr<media::AudioInputController> audio_controller_; |
96 scoped_refptr<net::URLRequestContextGetter> context_getter_; | 100 scoped_refptr<net::URLRequestContextGetter> context_getter_; |
97 AudioEncoder::Codec codec_; | 101 AudioEncoder::Codec codec_; |
98 scoped_ptr<AudioEncoder> encoder_; | 102 scoped_ptr<AudioEncoder> encoder_; |
99 Endpointer endpointer_; | 103 Endpointer endpointer_; |
100 int num_samples_recorded_; | 104 int num_samples_recorded_; |
101 float audio_level_; | 105 float audio_level_; |
102 AudioManager* audio_manager_; | 106 AudioManager* audio_manager_; |
103 | 107 |
104 DISALLOW_COPY_AND_ASSIGN(SpeechRecognizerImpl); | 108 DISALLOW_COPY_AND_ASSIGN(SpeechRecognizerImpl); |
105 }; | 109 }; |
106 | 110 |
107 } // namespace speech | 111 } // namespace speech |
108 | 112 |
109 #endif // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_ | 113 #endif // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNIZER_IMPL_H_ |
OLD | NEW |