Sphinx4 - IllegalArgumentException

Question

我已经创建了运行Sphinx4所需的所有文件(语言模型、字典和声学模型)。但是当我在Eclipse中运行它时，抛出了以下异常：

00:16:12.707 INFO unitManager          CI Unit: AE
00:16:12.713 INFO unitManager          CI Unit: AH
00:16:12.714 INFO unitManager          CI Unit: B
00:16:12.714 INFO unitManager          CI Unit: EY
00:16:12.715 INFO unitManager          CI Unit: F
00:16:12.715 INFO unitManager          CI Unit: IY
00:16:12.716 INFO unitManager          CI Unit: JH
00:16:12.716 INFO unitManager          CI Unit: L
00:16:12.717 INFO unitManager          CI Unit: M
00:16:12.722 INFO autoCepstrum         Cepstrum component auto-configured as follows: autoCepstrum {MelFrequencyFilterBank, DiscreteCosineTransform}
00:16:12.853 INFO dictionary           Loading dictionary from: file:Alphabets/tutorial/alphabets/etc/alphabets.dic
00:16:12.853 INFO dictionary           Loading filler dictionary from: file:Alphabets/tutorial/alphabets/model_parameters/alphabets.ci_cont/noisedict
00:16:12.854 INFO acousticModelLoader  Loading tied-state acoustic model from: file:Alphabets/tutorial/alphabets/model_parameters/alphabets.ci_cont
00:16:12.854 INFO acousticModelLoader  Pool means Entries: 30
00:16:12.855 INFO acousticModelLoader  Pool variances Entries: 30
00:16:12.855 INFO acousticModelLoader  Pool transition_matrices Entries: 10
00:16:12.855 INFO acousticModelLoader  Pool senones Entries: 30
00:16:12.855 INFO acousticModelLoader  Pool mixture_weights Entries: 30
00:16:12.856 INFO acousticModelLoader  Pool senones Entries: 30
00:16:12.856 INFO acousticModelLoader  Context Independent Unit Entries: 10
00:16:12.856 INFO acousticModelLoader  HMM Manager: 10 hmms
00:16:12.860 INFO acousticModel        CompositeSenoneSequences: 0
00:16:12.861 INFO largeTrigramModel    Loading n-gram language model from: file:Alphabets/tutorial/alphabets/etc/alphabets.lm.dmp
00:16:12.867 INFO largeTrigramModel    1-grams: 3
00:16:12.867 INFO largeTrigramModel    2-grams: 1
00:16:12.867 INFO largeTrigramModel    3-grams: 1
00:16:13.094 INFO lexTreeLinguist      Max CI Units 11
00:16:13.095 INFO lexTreeLinguist      Unit table size 1331
Exception in thread "main" java.lang.IllegalArgumentException
    at com.google.common.base.Preconditions.checkArgument(Preconditions.java:111)
    at edu.cmu.sphinx.linguist.WordSequence.getWord(WordSequence.java:179)
    at edu.cmu.sphinx.linguist.language.ngram.large.LargeNGramModel.getNGramProbDepth(LargeNGramModel.java:409)
    at edu.cmu.sphinx.linguist.language.ngram.large.LargeNGramModel.getNGramProbDepth(LargeNGramModel.java:412)
    at edu.cmu.sphinx.linguist.language.ngram.large.LargeNGramModel.getNGramProbDepth(LargeNGramModel.java:412)
    at edu.cmu.sphinx.linguist.language.ngram.large.LargeNGramModel.getProbDepth(LargeNGramModel.java:393)
    at edu.cmu.sphinx.linguist.lextree.LexTreeLinguist$LexTreeState.createWordStateArc(LexTreeLinguist.java:720)
    at edu.cmu.sphinx.linguist.lextree.LexTreeLinguist$LexTreeWordState.getSuccessors(LexTreeLinguist.java:1491)
    at edu.cmu.sphinx.decoder.search.WordPruningBreadthFirstSearchManager.collectSuccessorTokens(WordPruningBreadthFirstSearchManager.java:635)
    at edu.cmu.sphinx.decoder.search.WordPruningBreadthFirstSearchManager.growBranches(WordPruningBreadthFirstSearchManager.java:387)
    at edu.cmu.sphinx.decoder.search.WordPruningBreadthFirstSearchManager.localStart(WordPruningBreadthFirstSearchManager.java:359)
    at edu.cmu.sphinx.decoder.search.WordPruningBreadthFirstSearchManager.startRecognition(WordPruningBreadthFirstSearchManager.java:262)
    at edu.cmu.sphinx.decoder.Decoder.decode(Decoder.java:62)
    at edu.cmu.sphinx.recognizer.Recognizer.recognize(Recognizer.java:109)
    at edu.cmu.sphinx.recognizer.Recognizer.recognize(Recognizer.java:125)
    at edu.cmu.sphinx.api.AbstractSpeechRecognizer.getResult(AbstractSpeechRecognizer.java:50)
    at Main.main(Main.java:30)

这是我正在运行的程序，正如官方网站上所说的：

import java.io.IOException;
import java.util.Scanner;

import edu.cmu.sphinx.api.Configuration;
import edu.cmu.sphinx.api.LiveSpeechRecognizer;
import edu.cmu.sphinx.api.SpeechResult;

public class Main {

    public static void main(String[] args) {

        Configuration configuration = new Configuration();

        configuration
                .setAcousticModelPath("Alphabets/tutorial/alphabets/model_parameters/alphabets.ci_cont");

        configuration.setDictionaryPath("Alphabets/tutorial/alphabets/etc/alphabets.dic");

        configuration
                .setLanguageModelPath("Alphabets/tutorial/alphabets/etc/alphabets.lm.dmp");

        LiveSpeechRecognizer recognizer = null;
        try {
            recognizer = new LiveSpeechRecognizer(configuration);
        } catch (IOException e) {
            e.printStackTrace();
        }
        recognizer.startRecognition(true);

        SpeechResult result = recognizer.getResult();

        recognizer.stopRecognition();

        System.out.println(result.getHypothesis());
        result.getLattice().dumpDot("lattice.dot", "lattice");

    }
}

非常感谢您的帮助！

Answer 1

您的语言模型/Alphabets/tutorial/alphabets/etc/alphabets.lm.dmp是text arpa格式，但您向其添加了dmp扩展名。此手动编辑会混淆识别器。要解决此问题，请将alphabets.lm.dmp重命名为不带dmp扩展名的alphabets.lm，并在代码中编辑该名称。只需使用

configuration.setLanguageModelPath("Alphabets/tutorial/alphabets/etc/alphabets.lm");

你也没有足够的数据来训练模型，你的模型是不会工作的。有大量的数据用于训练是必须的。您可以在声学模型培训教程中找到详细信息

http://cmusphinx.sourceforge.net/wiki/tutorialam