基于决策树的藏语拉萨话三音子模型

Triphone models of Lhasa Tibetan based on decision tree

对藏语拉萨话中单音子及三音子分布情况进行了统计,分析了在藏语大词表连续词表连续语音识别中建立上下文相关声学模型的必要性。选择音素为建模单元,根据藏语特点,建立以音节为单位的发音字典。讨论了利用决策树建立三音子模型的几个关键问题和基本算法,结合国际音标分类和经验知识,确定了38个藏语拉萨话音子类别集及相应的决策树问题集。建立了共20个发音人8 170句的训练语料,在HTK平台上建立和训练得到了基于决策树的藏语拉萨话三音子模型,并分析了不同隐马尔可夫模型状态数及高斯混合度下的识别结果,确定了一套藏语大词表连续语音识别的完整方案。

Abstract：Probability distribution of monophones and triphones in Lhasa Tibetan are calculated and the necessity of establishing a contextual acoustic model in ASR for Lhasa Tibetan is analyzed. Phoneme is chosen as basic unit for acoustic models. According to the characteristics of Tibetan, a pronunciation dictionary based on single syllable is established. Main issues and algorithms for triphone models based on decision tree are discussed. According to IPAs and characteristics of Lhasa dialect, 38 phoneme sub- sets and question sets for triphone modeling are established. 8170 sentences of 20 speakers are recorded to train the models. Contextual continuous Hidden Markov Models（HMM） based on triphones are es tablished and trained on HTK platform. The recognition results under different sates number and mix tures are analyzed. And the framework for large-vocabulary continuous speech recognition of Lhasa Dia- lect is established.