Firstly, I introduce a kind of Context Dependent HMM model?Triphone, which cares about the neighboring context information of each phoneme.

• 首先提出了一种上下文相关模型-三音素模型，然后对这种模型进行优化。