论文部分内容阅读
具有确定词表的词组语音识别是语音识别研究的一个重要方面，应用相当广泛。本文在简单介绍词组语音识别之后给出了一个基于连续语音识别算法和词树约束的汉语词组语音识别方法。这种方法通过在束搜索连续语音识别算法中引入词树约束信息，发挥了连续语音识别算法的优点，并且充分利用了确定词表的约束信息，提高了计算和搜索的效率。然后介绍了约束词树和它的高效存储结构，这种结构提高了约束词树的存储效率和在识别搜索中的检索效率；最后给出实验的结果和讨论并进行简要的总结。
Phrase recognition with definite vocabulary Speech recognition is an important aspect of speech recognition and has a wide range of applications. This paper gives a speech recognition method based on continuous speech recognition algorithm and word tree constraint after Chinese speech recognition. In this method, the advantages of the continuous speech recognition algorithm are exerted by introducing the word tree constraint information into the continuous speech recognition algorithm of beam search, and the constraint information of the definite vocabularies is fully used to improve the efficiency of computation and search. Then, the constraint tree and its efficient storage structure are introduced. This structure improves the storage efficiency of the constraint tree and the retrieval efficiency in recognition search. Finally, the experimental results and discussions are given and a brief summary is given.