4

我会先说我在 C# 编程方面非常新手。我正在开发一个应用程序，用于使用 C# 和 SAPI v5.4 ( speechlib ) 以编程方式修改 Windows 语音词典。到目前为止一切都运行良好，但我需要更深入地了解字符串在合成（发声）时是如何解释的。

我的理解是，在 SAPI 5.4 中，单词被分解为音素表示，并且我在使用音素正确“训练”单词发音方面取得了一些成功。我也知道我可以手动将单词添加到 Windows 语音识别词典中，提供录音，然后提取单词的发音（音素）……但这很麻烦。探索默认情况下如何合成单词也很有用，即在没有我输入的情况下（例如合成器如何解释“海豚”？）。

从编码的角度来看，这是我到目前为止所得到的：

using System;
using System.Speech.Synthesis;

namespace SpeechTest
{
    class Program
    {
        static void Main(string[] args)
        {
            // Set up the speech synthesizer
            SpeechSynthesizer synthesizer = new SpeechSynthesizer();
            synthesizer.Volume = 100;
            synthesizer.Rate = -2;

            // Configure the audio output 
            synthesizer.SetOutputToDefaultAudioDevice();

            // Initialize string to store word of interest (not in the speech dictionary)
            string myWord = "dolphins";

            // Speak the word of interest
            synthesizer.Speak(myWord);

            // Retrieve pronunciation of myWord
            string myPronunciation = // *some code here*

            Console.WriteLine("Press any key to exit...");
            Console.ReadLine();
        }
    }
}
4

1 回答 1

1

Thanks to the amazing work of Casey Chesnut I've figured out how to determine the IPA phones for a given string. Now I just have to figure out how to convert from IPA phones to SAPI symbols, but that's for a separate topic (see here for how to get the SAPI phonemes from a text string). 

using System;
using System.Collections.ObjectModel;
using System.ComponentModel;
using System.IO;
using System.Speech.Recognition;
using System.Speech.Synthesis;
using System.Windows.Forms;

namespace SpeechTest
{
    class Program
    {
        static void Main(string[] args)
        {
            string MyText = "dolphins"; // Initialze string for storing word (or words) of interest
            string MyPronunciation = GetPronunciationFromText(MyText.Trim()); // Get IPA pronunciations of MyTe
            MessageBox.Show(MyText + " = " + MyPronunciation); // Output MyText and MyPronunciation
        }

        public static string recoPhonemes;

        public static string GetPronunciationFromText(string MyWord)
        {
            //this is a trick to figure out phonemes used by synthesis engine

            //txt to wav
            using (MemoryStream audioStream = new MemoryStream())
            {
                using (SpeechSynthesizer synth = new SpeechSynthesizer())
                {
                    synth.SetOutputToWaveStream(audioStream);
                    PromptBuilder pb = new PromptBuilder();
                    //pb.AppendBreak(PromptBreak.ExtraSmall); //'e' wont be recognized if this is large, or non-existent?
                    //synth.Speak(pb);
                    synth.Speak(MyWord);
                    //synth.Speak(pb);
                    synth.SetOutputToNull();
                    audioStream.Position = 0;

                    //now wav to txt (for reco phonemes)
                    recoPhonemes = String.Empty;
                    GrammarBuilder gb = new GrammarBuilder(MyWord);
                    Grammar g = new Grammar(gb); //TODO the hard letters to recognize are 'g' and 'e'
                    SpeechRecognitionEngine reco = new SpeechRecognitionEngine();
                    reco.SpeechHypothesized += new EventHandler<SpeechHypothesizedEventArgs>(reco_SpeechHypothesized);
                    reco.SpeechRecognitionRejected += new EventHandler<SpeechRecognitionRejectedEventArgs>(reco_SpeechRecognitionRejected);
                    reco.UnloadAllGrammars(); //only use the one word grammar
                    reco.LoadGrammar(g);
                    reco.SetInputToWaveStream(audioStream);
                    RecognitionResult rr = reco.Recognize();
                    reco.SetInputToNull();
                    if (rr != null)
                    {
                        recoPhonemes = StringFromWordArray(rr.Words, WordType.Pronunciation);
                    }
                    //txtRecoPho.Text = recoPhonemes;
                    return recoPhonemes;
                }
            }
        }

        public static string StringFromWordArray(ReadOnlyCollection<RecognizedWordUnit> words, WordType type)
        {
            string text = "";
            foreach (RecognizedWordUnit word in words)
            {
                string wordText = "";
                if (type == WordType.Text || type == WordType.Normalized)
                {
                    wordText = word.Text;
                }
                else if (type == WordType.Lexical)
                {
                    wordText = word.LexicalForm;
                }
                else if (type == WordType.Pronunciation)
                {
                    wordText = word.Pronunciation;
                    //MessageBox.Show(word.LexicalForm);
                }
                else
                {
                    throw new InvalidEnumArgumentException(String.Format("[0}: is not a valid input", type));
                }
                //Use display attribute

                if ((word.DisplayAttributes & DisplayAttributes.OneTrailingSpace) != 0)
                {
                    wordText += " ";
                }
                if ((word.DisplayAttributes & DisplayAttributes.TwoTrailingSpaces) != 0)
                {
                    wordText += "  ";
                }
                if ((word.DisplayAttributes & DisplayAttributes.ConsumeLeadingSpaces) != 0)
                {
                    wordText = wordText.TrimStart();
                }
                if ((word.DisplayAttributes & DisplayAttributes.ZeroTrailingSpaces) != 0)
                {
                    wordText = wordText.TrimEnd();
                }

                text += wordText;

            }
            return text;
        }

        public static void reco_SpeechHypothesized(object sender, SpeechHypothesizedEventArgs e)
        {
            recoPhonemes = StringFromWordArray(e.Result.Words, WordType.Pronunciation);
        }

        public static void reco_SpeechRecognitionRejected(object sender, SpeechRecognitionRejectedEventArgs e)
        {
            recoPhonemes = StringFromWordArray(e.Result.Words, WordType.Pronunciation);
        }

    }

    public enum WordType
    {
        Text,
        Normalized = Text,
        Lexical,
        Pronunciation
    }
}

// Credit for method of retrieving IPA pronunciation from a string goes to Casey Chesnut (http://www.mperfect.net/speechSamples/)
于 2018-03-27T20:22:08.837 回答