C#调用微软库的实现语音识别

早期，我们用SpeechRecognitionEngine如下来进行语音识别，但是复杂的词条加上环境噪声，其实成功率是不高的。

cs 复制代码

            SpeechRecognitionEngine recognizer = new SpeechRecognitionEngine(new System.Globalization.CultureInfo("zh-CN"));//初始化中文引擎
            Choices conmmonds = new Choices();
            conmmonds.Add(new string[] { "启动", "停止", "充电" });//添加词条
            GrammarBuilder gBuilder = new GrammarBuilder();
            gBuilder.Append(conmmonds);
            Grammar grammar = new DictationGrammar(); //new Grammar(gBuilder);
            grammar.Name = "default dictation";
            grammar.Enabled = true;
            recognizer.LoadGrammar(grammar);

            recognizer.SpeechRecognized += new EventHandler<SpeechRecognizedEventArgs>(Deal_SpeechRecongized);
            recognizer.SetInputToDefaultAudioDevice();//设置默认输入设备

            recognizer.InitialSilenceTimeout = TimeSpan.FromSeconds(3);
            recognizer.BabbleTimeout = TimeSpan.FromSeconds(2);
            recognizer.EndSilenceTimeout = TimeSpan.FromSeconds(1);
            recognizer.EndSilenceTimeoutAmbiguous = TimeSpan.FromSeconds(1.5);

            recognizer.RecognizeAsync(RecognizeMode.Multiple);//开始监控设备输入准备识别

            Speak(conmmonds + "测试语音");

另外一种方式，Microsoft的Speech SDK来实现语音的实时解读，然后根据接收到的文本内容进行处理，实现与用户的语音交互。

开发语音识别的需要使用注册Azure账号或测试api-key：

1. 准备工作

安装SDK或库 ：根据所选服务，下载并安装对应的C# SDK或NuGet包。比如，如果您选择Microsoft的Azure Cognitive Services中的语音服务，您就需要安装Microsoft.CognitiveServices.Speech NuGet包。

2. 实现语音识别

使用C#代码初始化语音识别客户端，监听麦克风输入，转换为文本。

cs 复制代码

1using Microsoft.CognitiveServices.Speech;
2using Microsoft.CognitiveServices.Speech.Audio;
3
4// 初始化SpeechConfig，使用您的API Key和区域
5var config = SpeechConfig.FromSubscription("your-api-key", "your-region");
6config.SpeechRecognitionLanguage = "zh-CN"; // 设置语言
7
8// 设置音频配置，使用默认麦克风
9using var audioConfig = AudioConfig.FromDefaultMicrophoneInput();
10using var recognizer = new SpeechRecognizer(config, audioConfig);
11
12Console.WriteLine("请说话...");
13
14// 开始连续语音识别
15var result = await recognizer.RecognizeOnceAsync();
16
17if (result.Reason == ResultReason.RecognizedSpeech)
18{
19    Console.WriteLine($"识别到的文本: {result.Text}");
20    // 在这里处理识别到的文本，比如调用另一个函数进行反馈处理
21}
22else if (result.Reason == ResultReason.NoMatch)
23{
24    Console.WriteLine("没有识别到有效语音输入。");
25}
26else if (result.Reason == ResultReason.Canceled)
27{
28    var cancellation = CancellationDetails.FromResult(result);
29    Console.WriteLine($"识别被取消: {cancellation.Reason}");
30}

3. 实现语音反馈

处理完识别到的文本后，可以使用相同的Speech SDK将处理结果转换为语音反馈给用户。

复制代码

1// 使用Text-to-Speech转换文本为语音
2var speechSynthesizer = new SpeechSynthesizer(config, null);
3
4var ssml = $"<speak version='1.0' xml:lang='zh-CN'><voice name='zh-CN-XiaoyiNeural'><prosody rate='medium'>{result.Text}</prosody></voice></speak>";
5var resultSynthesis = await speechSynthesizer.SynthesizeSsmlToWaveFileAsync(ssml, "output.wav");
6
7if (resultSynthesis.Reason == ResultReason.SynthesizingAudioCompleted)
8{
9    Console.WriteLine("语音合成完成。");
10    // 可以进一步处理，比如播放输出的音频文件
11}
12else if (resultSynthesis.Reason == ResultReason.Canceled)
13{
14    var cancellation = CancellationDetails.FromResult(resultSynthesis);
15    Console.WriteLine($"语音合成被取消: {cancellation.Reason}");
16}