基于隐马尔可夫模型的维吾尔语连续语音识别系统  

Uyghur continuous speech recognition system based on HMM

作　　者：那斯尔江.吐尔逊 吾守尔.斯拉木 

机构地区：[1]新疆大学数学与系统科学学院,乌鲁木齐830046 [2]西安交通大学电子与信息工程学院,西安710049 [3]新疆大学信息科学与工程学院,乌鲁木齐830046

出　　处：《计算机应用》2009年第7期2009-2011,2025,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(6076200660863008);国家语委重点项目(MZ115-75)

摘　　要：维吾尔语是黏着性语言,利用丰富的词缀可以用同样的词干产生超大词汇,给维吾尔语语音识别的研究工作带来了很大困难。结合维吾尔语自身特点,建立了维吾尔语连续语音语料库,利用HTK(HMMToolK it)工具实现了基于隐马尔可夫模型(HMM)的维吾尔语连续语音识别系统。在声学层,选取三音子作为基本的识别单元,建立了维吾尔语的三音子声学模型,并使用决策树、三音子绑定、修补哑音、增加高斯混合分量等方法提高模型的识别精度。在语言层,使用了适合于维吾尔语语音特征的基于统计的二元文法语言模型。最后,利用该系统进行了维吾尔语连续语音识别实验。Uyghur language is an agglutinative language. It is possible to produce a very large number of words from the same root with suffixes, so that the speech recognition of Uyghur language is very difficult. Combined with the characteristics of Uyghur language, this paper built a Uyghur continuous speech database, and designed the Hidden Markov Model （HMM） based Uyghur continuous speech recognition system by using the HTK （HMMToolKit）. On the acoustic level, this paper selected triphone as the basic recognition unit, and used many methods such as decision tree, tied-state triphones, fixing the silence models, increasing Gaussian mixture distribution to improve the precision of the models. On the language level, this study used the statistics-based bigram language model. Finally, this paper presented some recognition experiments by using that system.

关 键 词：维吾尔语 连续语音 语音识别 HMM的工具包 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]