微軟認知服務實現語音識別功能
想實現語音識別已經很久了，也嘗試了許多次，終究還是失敗了，原因很多，識別效果不理想，個人在技術上沒有成功實現，種種原因，以至於花費了好多時間在上面。語音識別，我嘗試過的有科大訊飛、百度語音，微軟系。最終還是喜歡微軟系的簡潔高效。(勿噴，純個人感覺)
最開始自己的想法是我說一句話(暫且在控制台上做Demo)，控制台程序能識別我說的是什么，然后顯示出來，並且根據我說的信息，執行相應的行為.(想法很美好，現實很糟心）初入語音識別，各種錯誤各種來，徘徊不定的選擇哪家公司的api，百度上查找各種語音識別的demo，學習參考，可是真正在.NET平台上運行成功的卻是寥寥無幾，或許是我查找方向有問題，經歷了許多的坑，沒一次成功過，心灰且意冷，打了幾次退堂鼓，卻終究忍受不住想玩語音識別。
可以看看我VS中的語音demo
第一個是今天的主角-稍后再提。
第二個和第三個是微軟系的系統自帶的System.Speech.dll和看了微軟博客里面的一篇文章而去嘗試的Microsoft.Speech.dll 可惜文章寫的挺好的，我嘗試卻是失敗 的，並且發現一個問題，就是英文版的微軟語音識別是無效的(Microsoft.Speech.Recognition)，而中文版的語音合成是無效的(Microsoft.Speech.Synthesis).，因 此，我不得不將兩個dll混合使用，來達到我想要的效果，最終效果確實達到了，不過卻是極其簡單的，一旦識別詞匯多起來，這識別率直接下降，我一直認為是采樣 頻率的問題，可是怎么也找不到采樣頻率的屬性或是字段，如有會的朋友可給我點信息，讓我也飛起來，哈哈。
第四個是百度語音識別demo，代碼簡潔許多，實現難度不難，可是小細節很多，需要注意，然后是雷區挺多的，但是呢，指導走出雷區的說明書卻是太少了，我是 踩了雷，很痛的那群。
首先來看看，現在市面上主流語音識別設計方式：
1、離線語音識別
離線語音識別很好理解，就是語音識別庫在本地或是局域網內，無需發起遠程連接。這個也是我當初的想法，自己弄一套語音識別庫，然后根據里面的內容設計想要的行為請求。利用微軟系的System.Speech.dll中的語音識別和語音合成功能。實現了簡單的中文語音識別功能，但是一旦我將語音識別庫逐漸加大，識別率就越來越低，不知是我電腦麥克風不行還是其它原因。最終受打擊，放棄。當我試着學習百度語音時，也發現了離線語音識別庫，但是呢官方並沒有給出具體的操作流程和設計思路，我也沒有去深入了解，有時間我要好好了解一番。
1 using System;
2 //using Microsoft.Speech.Synthesis;//中文版tts不能發聲
3 using Microsoft.Speech.Recognition;
4 using System.Speech.Synthesis;
5 //using System.Speech.Recognition;
6
7 namespace SAssassin.SpeechDemo
8 {
9 /// <summary>
10 /// 微軟語音識別 中文版 貌似效果還好點
11 /// </summary>
12 class Program
13 {
14 static SpeechSynthesizer sy = new SpeechSynthesizer();
15 static void Main(string[] args)
16 {
17 //創建中文識別器
18 using (SpeechRecognitionEngine recognizer = new SpeechRecognitionEngine(new System.Globalization.CultureInfo("zh-CN")))
19 {
20 foreach (var config in SpeechRecognitionEngine.InstalledRecognizers())
21 {
22 Console.WriteLine(config.Id);
23 }
24 //初始化命令詞
25 Choices commonds = new Choices();
26 string[] commond1 = new string[] { "一", "二", "三", "四", "五", "六", "七", "八", "九" };
27 string[] commond2 = new string[] { "很高興見到你", "識別率", "assassin", "長沙", "湖南", "實習" };
28 string[] commond3 = new string[] { "開燈", "關燈", "播放音樂", "關閉音樂", "澆水", "停止澆水", "打開背景燈", "關閉背景燈" };
29 //添加命令詞
30 commonds.Add(commond1);
31 commonds.Add(commond2);
32 commonds.Add(commond3);
33 //初始化命令詞管理
34 GrammarBuilder gBuilder = new GrammarBuilder();
35 //將命令詞添加到管理中
36 gBuilder.Append(commonds);
37 //實例化命令詞管理
38 Grammar grammar = new Grammar(gBuilder);
39
40 //創建並加載聽寫語法(添加命令詞匯識別的比較精准)
41 recognizer.LoadGrammarAsync(grammar);
42 //為語音識別事件添加處理程序。
43 recognizer.SpeechRecognized += new EventHandler<SpeechRecognizedEventArgs>(Recognizer_SpeechRRecongized);
44 //將輸入配置到語音識別器。
45 recognizer.SetInputToDefaultAudioDevice();
46 //啟動異步，連續語音識別。
47 recognizer.RecognizeAsync(RecognizeMode.Multiple);
48 //保持控制台窗口打開。
49 Console.WriteLine("你好");
50 sy.Speak("你好");
51 Console.ReadLine();
52 }
53 }
54
55 //speechrecognized事件處理
56 static void Recognizer_SpeechRRecongized(object sender, SpeechRecognizedEventArgs e)
57 {
58 Console.WriteLine("識別結果：" + e.Result.Text + " " + e.Result.Confidence + " " + DateTime.Now);
59 sy.Speak(e.Result.Text);
60 }
61 }
62 }
2、在線語音識別。
在線語音識別是我們當前程序將語音文件發送到遠程服務中心，待遠程服務中心匹配解決后將匹配結果進行返回的過程。其使用的一般是Restful風格，利用Json數據往返識別結果。
剛開始學習科大訊飛的語音識別，剛開始什么也不懂，聽朋友推薦加上自己百度學習，科大訊飛都說很不錯，也抱着心態去學習學習，可是windows平台下只有C++的demo，無奈我是C#，雖說語言很大程度上不分家，可是不想過於麻煩，網上找了一個demo，據說是最全的C#版本的訊飛語音識別demo,可是當看到里面錯綜復雜的源代碼時，內心是憂傷的，這里是直接通過一種方式引用c++的函數，運行了該demo，成功了，能簡單的錄音然后識別，但是有些地方存在問題，也得不到解決方案，不得已，放棄。
后來，百度語音吸引我了，七月份時，重新開始看百度語音的demo，官網demo比較簡單，嘗試着學習了一下，首先你得到百度語音開放平台去創建應用得到App key 和Secret key，然后下載着demo，在構造函數或者字段中又或是寫入配置文件中，將這兩個得到的key寫入，程序會根據這兩個key去發起請求的。就如同開頭所說，這是在線語音識別，利用Restful風格，將語音文件上傳至百度語音識別中心，然后識別后將回執數據返回到我們的程序中，剛開始，配置的時候自己技術不怎么樣，配置各種出錯，地雷開始踩了，總要炸幾次，最終還是能將demo中的測試文件識別出來，算是我個人的一小步把.(如果有朋友正好碰到踩雷問題，不妨可與我一起探討，或許我也不懂，但在我踩過的里面至少我懂了，哈哈)
接下來是設計思路的問題，語音識別能成功了，語音合成也能成功了，這里要注意，語音識別和語音合成要分別開通，並且這兩個都有App Key和Secret Key 雖然是一樣的，但是還是要注意，不然語音合成就會出問題的。接下來要考慮的問題就是，百度語音的設計思路是根據文件識別，但是我們考慮的最多的就是我直接麥克風語音輸入，然后識別，這也是我的想法，接下來解決這一問題，設計思路是，我將輸入的信息作為文件形式保存，等我輸入完，然后就調用語音識別方法，這不就行了嗎，確實也是可以的，此處，又開始進入雷區了，利用NAudio.dll文件實現錄音功能，這個包可以在Nuget中下載。
1 using NAudio.Wave;
2 using System;
3
4 namespace SAssassin.VOC
5 {
6 /// <summary>
7 /// 實現錄音功能
8 /// </summary>
9 public class RecordWaveToFile
10 {
11 private WaveFileWriter waveFileWriter = null;
12 private WaveIn myWaveIn = null;
13
14 public void StartRecord()
15 {
16 ConfigWave();
17 myWaveIn.StartRecording();
18 }
19
20 private void ConfigWave()
21 {
22 string filePath = AppDomain.CurrentDomain.BaseDirectory + "Temp.wav";
23 myWaveIn = new WaveIn()
24 {
25 WaveFormat = new WaveFormat(16000, 16, 1)//8k,16bit,單頻
26 //WaveFormat = new WaveFormat()//識別音質清晰
27 };
28 myWaveIn.DataAvailable += new System.EventHandler<WaveInEventArgs>(WaveIn_DataAvailable);
29 myWaveIn.RecordingStopped += new System.EventHandler<StoppedEventArgs>(WaveIn_RecordingStopped);
30 waveFileWriter = new WaveFileWriter(filePath, myWaveIn.WaveFormat);
31 }
32
33 private void WaveIn_DataAvailable(object sender,WaveInEventArgs e)
34 {
35 if(waveFileWriter != null)
36 {
37 waveFileWriter.Write(e.Buffer,0,e.BytesRecorded);
38 waveFileWriter.Flush();
39 }
40 }
41
42 private void WaveIn_RecordingStopped(object sender,StoppedEventArgs e)
43 {
44 myWaveIn.StopRecording();
45 }
46 }
47 }
此處控制器中使用WaveInEvent不會報錯，可就在這之前，我用的是WaveIn類，然后直接報錯了
“System.InvalidOperationException:“Use WaveInEvent to record on a background thread””
在StackOverFlow上找到了解決方案，就是將WaveIn類換成WaveInEvent類即可，進入類里面看一下，其實發現都是引用同一個接口，甚至說兩個類的結構都是一模一樣的，只是一個用於GUI線程，一個用於后台線程。一切就緒，錄音也能實現，可是當我查看自己的錄音文件時，雜音好多，音質不侵襲，甚至是直接失真了，沒什么用，送百度也識別失敗，當將采樣頻率提高到44k時效果很好，錄音文件很不錯，但是問題來了，百度語音識別規定的pcm文件只能是8k-16bit,糟心，想換成其它格式的文件，采取壓縮形式保存，但是一旦將采樣頻率降下來，這個效果就很糟糕，識別也是成了問題。不得不說，這還要慢慢來解決哈。
進入今天重頭戲，這也是我博客園第一篇隨筆文章，該講點重點了，微軟認知服務，七月中旬的時候接觸到了必應的語音識別api，在微軟bing官網里，並且里面的識別效果，讓我驚呼，這識別率太高了。然后想找它的api，發現文檔全是英文資料，糟心。把資料看完，感覺使用方式很不錯，也是遠程調用的方式，但是api呢，官網找了老半天，只有文檔，那時也沒看上面的產品，試用版什么的，只能看着，卻不能用，心累。也就在這幾天，重新看了下必應的語音識別文檔，才接觸到這個詞--"微軟認知服務", 恕我見識太淺，這個好東西卻沒聽過，百度一查，真是不錯，微軟太牛了，這個里面包含很多api，語音識別都只算小菜一只，人臉識別，語義感知，等等很牛的功能，找到Api，找到免費試用，登錄獲得app的secret key ，便可以用起來了。下載一個demo，將secret key輸入，測試一下，哇塞，這識別效果，簡直了，太強了。並且從百度中看到很多結果，使用到微軟認知服務語音識別功能的很少，我也因此有寫一點東西的想法。
我將demo中的很多地方抽出來直接形成了一個控制器程序，源碼如下
1 public class SpeechConfig
2 {
3 #region Fields
4 /// <summary>
5 /// The isolated storage subscription key file name.
6 /// </summary>
7 private const string IsolatedStorageSubscriptionKeyFileName = "Subscription.txt";
8
9 /// <summary>
10 /// The default subscription key prompt message
11 /// </summary>
12 private const string DefaultSubscriptionKeyPromptMessage = "Secret key";
13
14 /// <summary>
15 /// You can also put the primary key in app.config, instead of using UI.
16 /// string subscriptionKey = ConfigurationManager.AppSettings["primaryKey"];
17 /// </summary>
18 private string subscriptionKey = ConfigurationManager.AppSettings["primaryKey"];
19
20 /// <summary>
21 /// Gets or sets subscription key
22 /// </summary>
23 public string SubscriptionKey
24 {
25 get
26 {
27 return this.subscriptionKey;
28 }
29
30 set
31 {
32 this.subscriptionKey = value;
33 this.OnPropertyChanged<string>();
34 }
35 }
36
37 /// <summary>
38 /// The data recognition client
39 /// </summary>
40 private DataRecognitionClient dataClient;
41
42 /// <summary>
43 /// The microphone client
44 /// </summary>
45 private MicrophoneRecognitionClient micClient;
46
47 #endregion Fields
48
49 #region event
50 /// <summary>
51 /// Implement INotifyPropertyChanged interface
52 /// </summary>
53 public event PropertyChangedEventHandler PropertyChanged;
54
55 /// <summary>
56 /// Helper function for INotifyPropertyChanged interface
57 /// </summary>
58 /// <typeparam name="T">Property type</typeparam>
59 /// <param name="caller">Property name</param>
60 private void OnPropertyChanged<T>([CallerMemberName]string caller = null)
61 {
62 this.PropertyChanged?.Invoke(this, new PropertyChangedEventArgs(caller));
63 }
64 #endregion event
65
66 #region 屬性
67 /// <summary>
68 /// Gets the current speech recognition mode.
69 /// </summary>
70 /// <value>
71 /// The speech recognition mode.
72 /// </value>
73 private SpeechRecognitionMode Mode
74 {
75 get
76 {
77 if (this.IsMicrophoneClientDictation ||
78 this.IsDataClientDictation)
79 {
80 return SpeechRecognitionMode.LongDictation;
81 }
82
83 return SpeechRecognitionMode.ShortPhrase;
84 }
85 }
86
87 /// <summary>
88 /// Gets the default locale.
89 /// </summary>
90 /// <value>
91 /// The default locale.
92 /// </value>
93 private string DefaultLocale
94 {
95 //get { return "en-US"; }
96 get { return "zh-CN"; }
97
98 }
99
100 /// <summary>
101 /// Gets the Cognitive Service Authentication Uri.
102 /// </summary>
103 /// <value>
104 /// The Cognitive Service Authentication Uri. Empty if the global default is to be used.
105 /// </value>
106 private string AuthenticationUri
107 {
108 get
109 {
110 return ConfigurationManager.AppSettings["AuthenticationUri"];
111 }
112 }
113
114 /// <summary>
115 /// Gets a value indicating whether or not to use the microphone.
116 /// </summary>
117 /// <value>
118 /// <c>true</c> if [use microphone]; otherwise, <c>false</c>.
119 /// </value>
120 private bool UseMicrophone
121 {
122 get
123 {
124 return this.IsMicrophoneClientWithIntent ||
125 this.IsMicrophoneClientDictation ||
126 this.IsMicrophoneClientShortPhrase;
127 }
128 }
129
130 /// <summary>
131 /// Gets the short wave file path.
132 /// </summary>
133 /// <value>
134 /// The short wave file.
135 /// </value>
136 private string ShortWaveFile
137 {
138 get
139 {
140 return ConfigurationManager.AppSettings["ShortWaveFile"];
141 }
142 }
143
144 /// <summary>
145 /// Gets the long wave file path.
146 /// </summary>
147 /// <value>
148 /// The long wave file.
149 /// </value>
150 private string LongWaveFile
151 {
152 get
153 {
154 return ConfigurationManager.AppSettings["LongWaveFile"];
155 }
156 }
157 #endregion 屬性
158
159 #region 模式選擇控制器設置
160 /// <summary>
161 /// Gets or sets a value indicating whether this instance is microphone client short phrase.
162 /// </summary>
163 /// <value>
164 /// <c>true</c> if this instance is microphone client short phrase; otherwise, <c>false</c>.
165 /// </value>
166 public bool IsMicrophoneClientShortPhrase { get; set; }
167
168 /// <summary>
169 /// Gets or sets a value indicating whether this instance is microphone client dictation.
170 /// </summary>
171 /// <value>
172 /// <c>true</c> if this instance is microphone client dictation; otherwise, <c>false</c>.
173 /// </value>
174 public bool IsMicrophoneClientDictation { get; set; }
175
176 /// <summary>
177 /// Gets or sets a value indicating whether this instance is microphone client with intent.
178 /// </summary>
179 /// <value>
180 /// <c>true</c> if this instance is microphone client with intent; otherwise, <c>false</c>.
181 /// </value>
182 public bool IsMicrophoneClientWithIntent { get; set; }
183
184 /// <summary>
185 /// Gets or sets a value indicating whether this instance is data client short phrase.
186 /// </summary>
187 /// <value>
188 /// <c>true</c> if this instance is data client short phrase; otherwise, <c>false</c>.
189 /// </value>
190 public bool IsDataClientShortPhrase { get; set; }
191
192 /// <summary>
193 /// Gets or sets a value indicating whether this instance is data client with intent.
194 /// </summary>
195 /// <value>
196 /// <c>true</c> if this instance is data client with intent; otherwise, <c>false</c>.
197 /// </value>
198 public bool IsDataClientWithIntent { get; set; }
199
200 /// <summary>
201 /// Gets or sets a value indicating whether this instance is data client dictation.
202 /// </summary>
203 /// <value>
204 /// <c>true</c> if this instance is data client dictation; otherwise, <c>false</c>.
205 /// </value>
206 public bool IsDataClientDictation { get; set; }
207
208 #endregion
209
210 #region 委托執行對象
211 /// <summary>
212 /// Called when the microphone status has changed.
213 /// </summary>
214 /// <param name="sender">The sender.</param>
215 /// <param name="e">The <see cref="MicrophoneEventArgs"/> instance containing the event data.</param>
216 private void OnMicrophoneStatus(object sender, MicrophoneEventArgs e)
217 {
218 Task task = new Task(() =>
219 {
220 Console.WriteLine("--- Microphone status change received by OnMicrophoneStatus() ---");
221 Console.WriteLine("********* Microphone status: {0} *********", e.Recording);
222 if (e.Recording)
223 {
224 Console.WriteLine("Please start speaking.");
225 }
226
227 Console.WriteLine();
228 });
229 task.Start();
230 }
231
232 /// <summary>
233 /// Called when a partial response is received.
234 /// </summary>
235 /// <param name="sender">The sender.</param>
236 /// <param name="e">The <see cref="PartialSpeechResponseEventArgs"/> instance containing the event data.</param>
237 private void OnPartialResponseReceivedHandler(object sender, PartialSpeechResponseEventArgs e)
238 {
239 Console.WriteLine("--- Partial result received by OnPartialResponseReceivedHandler() ---");
240 Console.WriteLine("{0}", e.PartialResult);
241 Console.WriteLine();
242 }
243
244 /// <summary>
245 /// Called when an error is received.
246 /// </summary>
247 /// <param name="sender">The sender.</param>
248 /// <param name="e">The <see cref="SpeechErrorEventArgs"/> instance containing the event data.</param>
249 private void OnConversationErrorHandler(object sender, SpeechErrorEventArgs e)
250 {
251 Console.WriteLine("--- Error received by OnConversationErrorHandler() ---");
252 Console.WriteLine("Error code: {0}", e.SpeechErrorCode.ToString());
253 Console.WriteLine("Error text: {0}", e.SpeechErrorText);
254 Console.WriteLine();
255 }
256
257 /// <summary>
258 /// Called when a final response is received;
259 /// </summary>
260 /// <param name="sender">The sender.</param>
261 /// <param name="e">The <see cref="SpeechResponseEventArgs"/> instance containing the event data.</param>
262 private void OnMicShortPhraseResponseReceivedHandler(object sender, SpeechResponseEventArgs e)
263 {
264 Task task = new Task(() =>
265 {
266 Console.WriteLine("--- OnMicShortPhraseResponseReceivedHandler ---");
267
268 // we got the final result, so it we can end the mic reco. No need to do this
269 // for dataReco, since we already called endAudio() on it as soon as we were done
270 // sending all the data.
271 this.micClient.EndMicAndRecognition();
272
273 this.WriteResponseResult(e);
274 });
275 task.Start();
276 }
277
278 /// <summary>
279 /// Called when a final response is received;
280 /// </summary>
281 /// <param name="sender">The sender.</param>
282 /// <param name="e">The <see cref="SpeechResponseEventArgs"/> instance containing the event data.</param>
283 private void OnDataShortPhraseResponseReceivedHandler(object sender, SpeechResponseEventArgs e)
284 {
285 Task task = new Task(() =>
286 {
287 Console.WriteLine("--- OnDataShortPhraseResponseReceivedHandler ---");
288
289 // we got the final result, so it we can end the mic reco. No need to do this
290 // for dataReco, since we already called endAudio() on it as soon as we were done
291 // sending all the data.
292 this.WriteResponseResult(e);
293
294 });
295 task.Start();
296 }
297
298 /// <summary>
299 /// Called when a final response is received;
300 /// </summary>
301 /// <param name="sender">The sender.</param>
302 /// <param name="e">The <see cref="SpeechResponseEventArgs"/> instance containing the event data.</param>
303 private void OnMicDictationResponseReceivedHandler(object sender, SpeechResponseEventArgs e)
304 {
305 Console.WriteLine("--- OnMicDictationResponseReceivedHandler ---");
306 if (e.PhraseResponse.RecognitionStatus == RecognitionStatus.EndOfDictation ||
307 e.PhraseResponse.RecognitionStatus == RecognitionStatus.DictationEndSilenceTimeout)
308 {
309 Task task = new Task(() =>
310 {
311 // we got the final result, so it we can end the mic reco. No need to do this
312 // for dataReco, since we already called endAudio() on it as soon as we were done
313 // sending all the data.
314 this.micClient.EndMicAndRecognition();
315 });
316 task.Start();
317 }
318
319 this.WriteResponseResult(e);
320 }
321
322 /// <summary>
323 /// Called when a final response is received;
324 /// </summary>
325 /// <param name="sender">The sender.</param>
326 /// <param name="e">The <see cref="SpeechResponseEventArgs"/> instance containing the event data.</param>
327 private void OnDataDictationResponseReceivedHandler(object sender, SpeechResponseEventArgs e)
328 {
329 Console.WriteLine("--- OnDataDictationResponseReceivedHandler ---");
330 if (e.PhraseResponse.RecognitionStatus == RecognitionStatus.EndOfDictation ||
331 e.PhraseResponse.RecognitionStatus == RecognitionStatus.DictationEndSilenceTimeout)
332 {
333 Task task = new Task(() =>
334 {
335
336 // we got the final result, so it we can end the mic reco. No need to do this
337 // for dataReco, since we already called endAudio() on it as soon as we were done
338 // sending all the data.
339 });
340 task.Start();
341 }
342
343 this.WriteResponseResult(e);
344 }
345
346 /// <summary>
347 /// Sends the audio helper.
348 /// </summary>
349 /// <param name="wavFileName">Name of the wav file.</param>
350 private void SendAudioHelper(string wavFileName)
351 {
352 using (FileStream fileStream = new FileStream(wavFileName, FileMode.Open, FileAccess.Read))
353 {
354 // Note for wave files, we can just send data from the file right to the server.
355 // In the case you are not an audio file in wave format, and instead you have just
356 // raw data (for example audio coming over bluetooth), then before sending up any
357 // audio data, you must first send up an SpeechAudioFormat descriptor to describe
358 // the layout and format of your raw audio data via DataRecognitionClient's sendAudioFormat() method.
359 int bytesRead = 0;
360 byte[] buffer = new byte[1024];
361
362 try
363 {
364 do
365 {
366 // Get more Audio data to send into byte buffer.
367 bytesRead = fileStream.Read(buffer, 0, buffer.Length);
368
369 // Send of audio data to service.
370 this.dataClient.SendAudio(buffer, bytesRead);
371 }
372 while (bytesRead > 0);
373 }
374 finally
375 {
376 // We are done sending audio. Final recognition results will arrive in OnResponseReceived event call.
377 this.dataClient.EndAudio();
378 }
379 }
380 }
381 #endregion 委托執行對象
382
383 #region 輔助方法
384 /// <summary>
385 /// Gets the subscription key from isolated storage.
386 /// </summary>
387 /// <returns>The subscription key.</returns>
388 private string GetSubscriptionKeyFromIsolatedStorage()
389 {
390 string subscriptionKey = null;
391
392 using (IsolatedStorageFile isoStore = IsolatedStorageFile.GetStore(IsolatedStorageScope.User | IsolatedStorageScope.Assembly, null, null))
393 {
394 try
395 {
396 using (var iStream = new IsolatedStorageFileStream(IsolatedStorageSubscriptionKeyFileName, FileMode.Open, isoStore))
397 {
398 using (var reader = new StreamReader(iStream))
399 {
400 subscriptionKey = reader.ReadLine();
401 }
402 }
403 }
404 catch (FileNotFoundException)
405 {
406 subscriptionKey = null;
407 }
408 }
409
410 if (string.IsNullOrEmpty(subscriptionKey))
411 {
412 subscriptionKey = DefaultSubscriptionKeyPromptMessage;
413 }
414
415 return subscriptionKey;
416 }
417
418 /// <summary>
419 /// Creates a new microphone reco client without LUIS intent support.
420 /// </summary>
421 private void CreateMicrophoneRecoClient()
422 {
423 this.micClient = SpeechRecognitionServiceFactory.CreateMicrophoneClient(
424 this.Mode,this.DefaultLocale,this.SubscriptionKey);
425
426 this.micClient.AuthenticationUri = this.AuthenticationUri;
427
428 // Event handlers for speech recognition results
429 this.micClient.OnMicrophoneStatus += this.OnMicrophoneStatus;
430 this.micClient.OnPartialResponseReceived += this.OnPartialResponseReceivedHandler;
431 if (this.Mode == SpeechRecognitionMode.ShortPhrase)
432 {
433 this.micClient.OnResponseReceived += this.OnMicShortPhraseResponseReceivedHandler;
434 }
435 else if (this.Mode == SpeechRecognitionMode.LongDictation)
436 {
437 this.micClient.OnResponseReceived += this.OnMicDictationResponseReceivedHandler;
438 }
439
440 this.micClient.OnConversationError += this.OnConversationErrorHandler;
441 }
442
443 /// <summary>
444 /// Creates a data client without LUIS intent support.
445 /// Speech recognition with data (for example from a file or audio source).
446 /// The data is broken up into buffers and each buffer is sent to the Speech Recognition Service.
447 /// No modification is done to the buffers, so the user can apply their
448 /// own Silence Detection if desired.
449 /// </summary>
450 private void CreateDataRecoClient()
451 {
452 this.dataClient = SpeechRecognitionServiceFactory.CreateDataClient(
453 this.Mode,
454 this.DefaultLocale,
455 this.SubscriptionKey);
456 this.dataClient.AuthenticationUri = this.AuthenticationUri;
457
458 // Event handlers for speech recognition results
459 if (this.Mode == SpeechRecognitionMode.ShortPhrase)
460 {
461 this.dataClient.OnResponseReceived += this.OnDataShortPhraseResponseReceivedHandler;
462 }
463 else
464 {
465 this.dataClient.OnResponseReceived += this.OnDataDictationResponseReceivedHandler;
466 }
467
468 this.dataClient.OnPartialResponseReceived += this.OnPartialResponseReceivedHandler;
469 this.dataClient.OnConversationError += this.OnConversationErrorHandler;
470 }
471
472 /// <summary>
473 /// Writes the response result.
474 /// </summary>
475 /// <param name="e">The <see cref="SpeechResponseEventArgs"/> instance containing the event data.</param>
476 private void WriteResponseResult(SpeechResponseEventArgs e)
477 {
478 if (e.PhraseResponse.Results.Length == 0)
479 {
480 Console.WriteLine("No phrase response is available.");
481 }
482 else
483 {
484 Console.WriteLine("********* Final n-BEST Results *********");
485 for (int i = 0; i < e.PhraseResponse.Results.Length; i++)
486 {
487 Console.WriteLine(
488 "[{0}] Confidence={1}, Text=\"{2}\"",
489 i,
490 e.PhraseResponse.Results[i].Confidence,
491 e.PhraseResponse.Results[i].DisplayText);
492 if (e.PhraseResponse.Results[i].DisplayText == "關閉。")
493 {
494 Console.WriteLine("收到命令，馬上關閉");
495 }
496 }
497
498 Console.WriteLine();
499 }
500 }
501 #endregion 輔助方法
502
503 #region Init
504 public SpeechConfig()
505 {
506 this.IsMicrophoneClientShortPhrase = true;
507 this.IsMicrophoneClientWithIntent = false;
508 this.IsMicrophoneClientDictation = false;
509 this.IsDataClientShortPhrase = false;
510 this.IsDataClientWithIntent = false;
511 this.IsDataClientDictation = false;
512
513 this.SubscriptionKey = this.GetSubscriptionKeyFromIsolatedStorage();
514 }
515
516 /// <summary>
517 /// 語音識別開始執行
518 /// </summary>
519 public void SpeechRecognize()
520 {
521 if (this.UseMicrophone)
522 {
523 if (this.micClient == null)
524 {
525 this.CreateMicrophoneRecoClient();
526 }
527
528 this.micClient.StartMicAndRecognition();
529 }
530 else
531 {
532 if (null == this.dataClient)
533 {
534 this.CreateDataRecoClient();
535 }
536
537 this.SendAudioHelper((this.Mode == SpeechRecognitionMode.ShortPhrase) ? this.ShortWaveFile : this.LongWaveFile);
538 }
539 }
540 #endregion Init
541 }
在這其中有幾個引用文件可以通過nuget包下載，基本沒什么問題。
對了這里注意的一個問題就是，下載Microsoft.Speech的時候一定是兩個包都需要下載，不然會報錯的，版本必須是4.5+以上的。
只需替換默認的key就行，程序便可跑起來，效果真是很6
這識別率真是很好很好，很滿意，可是這個微軟的免費試用只有一個月，那就只能在這個月里多讓它開花結果了哈哈。