管理音频输入问题

  • 项目

了解如何管理由音频输入质量所导致的语音识别准确度问题。

重要 API：SpeechRecognizer、RecognitionQualityDegrading、SpeechRecognitionAudioProblem

评估音频输入质量

当语音识别处于活动状态时，使用语音识别器的 RecognitionQualityDegrading 事件，确定一个或多个音频问题是否会干扰语音输入。 事件参数 (SpeechRecognitionQualityDegradingEventArgs) 可提供 Problem 属性，该属性描述了检测到的音频输入问题。

背景噪音太大、麦克风设为静音以及扬声器的音量或语速都可能会影响语音识别。

在此情况下，我们可以配置一个语音识别器并开始侦听 RecognitionQualityDegrading 事件。

private async void WeatherSearch_Click(object sender, RoutedEventArgs e)
{
    // Create an instance of SpeechRecognizer.
    var speechRecognizer = new Windows.Media.SpeechRecognition.SpeechRecognizer();

    // Listen for audio input issues.
    speechRecognizer.RecognitionQualityDegrading += speechRecognizer_RecognitionQualityDegrading;

    // Add a web search grammar to the recognizer.
    var webSearchGrammar = new Windows.Media.SpeechRecognition.SpeechRecognitionTopicConstraint(Windows.Media.SpeechRecognition.SpeechRecognitionScenario.WebSearch, "webSearch");


    speechRecognizer.UIOptions.AudiblePrompt = "Say what you want to search for...";
    speechRecognizer.UIOptions.ExampleText = "Ex. 'weather for London'";
    speechRecognizer.Constraints.Add(webSearchGrammar);

    // Compile the constraint.
    await speechRecognizer.CompileConstraintsAsync();

    // Start recognition.
    Windows.Media.SpeechRecognition.SpeechRecognitionResult speechRecognitionResult = await speechRecognizer.RecognizeWithUIAsync();
    //await speechRecognizer.RecognizeWithUIAsync();

    // Do something with the recognition result.
    var messageDialog = new Windows.UI.Popups.MessageDialog(speechRecognitionResult.Text, "Text spoken");
    await messageDialog.ShowAsync();
}

管理语音识别体验

使用由 Problem 属性提供的说明可帮助用户改善识别的条件。

在此处，我们将为 RecognitionQualityDegrading 事件创建一个处理程序，用于检查较低的音量级别。 然后，我们将使用 SpeechSynthesizer 对象建议用户尝试提高音量。

private async void speechRecognizer_RecognitionQualityDegrading(
    Windows.Media.SpeechRecognition.SpeechRecognizer sender,
    Windows.Media.SpeechRecognition.SpeechRecognitionQualityDegradingEventArgs args)
{
    // Create an instance of a speech synthesis engine (voice).
    var speechSynthesizer =
        new Windows.Media.SpeechSynthesis.SpeechSynthesizer();

    // If input speech is too quiet, prompt the user to speak louder.
    if (args.Problem == Windows.Media.SpeechRecognition.SpeechRecognitionAudioProblem.TooQuiet)
    {
        // Generate the audio stream from plain text.
        Windows.Media.SpeechSynthesis.SpeechSynthesisStream stream;
        try
        {
            stream = await speechSynthesizer.SynthesizeTextToStreamAsync("Try speaking louder");
            stream.Seek(0);
        }
        catch (Exception)
        {
            stream = null;
        }

        // Send the stream to the MediaElement declared in XAML.
        await CoreApplication.MainView.CoreWindow.Dispatcher.RunAsync(Windows.UI.Core.CoreDispatcherPriority.High, () =>
        {
            this.media.SetSource(stream, stream.ContentType);
        });
    }
}

示例