SpeechRecognitionEngine 识别语法中没有的单词

Question

我正在尝试使用SpeechRecognitionEngine 构建语音识别程序。我想要实现的功能之一是我可以google 我所说的短语。我应该说：例如“谷歌+我的短语”。但问题是SpeechRecognitionEngine 只能识别我将在Grammar 中添加的单词，所以它无法识别我的短语。我该如何实现这个功能？

Answer 1

这是可能的，但更难实现。基本上，您将系统置于持续聆听/听写模式，并为每个识别的短语选择要做什么。

找到了您可能也想尝试的答案： C# SAPI - Recognizing phrases without pre-defined condition statements

这是一个简单的例子（这是.net 4.5，WinForms），希望对您有所帮助

using System;
using System.Windows.Forms;
using System.Speech.Recognition;
using System.Speech.Synthesis;
using System.Diagnostics;
using System.Text.RegularExpressions;

public partial class Main : Form
{
    private SpeechRecognitionEngine listener;
    private SpeechSynthesizer speaker;      

    public Main()
    {
        InitializeComponent();
    }

    private void Main_Load(object sender, EventArgs e)
    {
        speaker = new SpeechSynthesizer();
        speaker.SelectVoice("Microsoft David Desktop");

        GrammarBuilder builder = new GrammarBuilder();
        builder.AppendDictation();

        Grammar grammar = new Grammar(builder);

        listener = new SpeechRecognitionEngine();
        listener.LoadGrammar(grammar);
        listener.SetInputToDefaultAudioDevice();
        listener.SpeechRecognized += new EventHandler<SpeechRecognizedEventArgs>(listener_SpeechRecognized);
        listener.RecognizeAsync(RecognizeMode.Multiple);
    }

    void listener_SpeechRecognized(object sender, SpeechRecognizedEventArgs e)
    {
        string commandName = e.Result.Text;

        switch (commandName.ToLower())
        {
            case "exit":

                speaker.Speak("talk to you later, bye");
                Application.Exit();
                break;

            case "stop speaking":
            case "stop talking":
            case "be quiet":
            case "silence":

                speaker.SpeakAsyncCancelAll();
                break;

            case "hello":

                speaker.SpeakAsync("hello, how are you doing?");
                break;

            case "i'm fine":

                speaker.SpeakAsync("i am glad to hear that");
                break;

            case "thank you":

                speaker.SpeakAsync("you're welcome");
                break;

            case "thanks":

                speaker.SpeakAsync("no problem");
                break;

            case "what's the time":
            case "what time is it":
            case "time":

                speaker.SpeakAsync(DateTime.Now.ToShortTimeString());

                break;

            case "what's the date":
            case "what day is it":
            case "what is today's date":
            case "what is today":
            case "today":

                speaker.SpeakAsync(DateTime.Today.ToString("dddd, MMMM d, yyyy"));

                break;

            case "do read it":

                Process.Start("http://www.reddit.com");
                break;

            case "do face book":

                Process.Start("http://www.facebook.com");
                break;

            case "do search":

                Process.Start("http://www.google.com");
                break;

            case "do videos":

                Process.Start("http://www.youtube.com");
                break;

            default:

                //handle non-normalized recognition
                Match m = Regex.Match(commandName, "YOUR_PATTERN_HERE");

                if (m.Success)
                {
                    speaker.SpeakAsync("I found a match");

                    //example, probably should URL encode the value...
                    //Process.Start("http://www.google.com?q=" + m.Value);
                }

                break;
        }
    }

    private void Main_FormClosing(object sender, FormClosingEventArgs e)
    {
        //be sure to clean up!
        listener.UnloadAllGrammars();
        listener.Dispose();
        listener = null;

        speaker.Dispose();
        speaker = null;

        grammar = null;
    }
}

Answer 2

这在语音技术的当前阶段是不可能的。你能做的最好的就是让语法足够大，这样你会觉得所有的单词都被识别了，但是准确性会大大降低。