In this thesis, we examine the techniques which can realize the
context-dependent phoneme modeling to model co-articulatory effects of korean
language, The context-independent phoneme models are constructed using 51
phoneme-like units. Interpolatio...
In this thesis, we examine the techniques which can realize the
context-dependent phoneme modeling to model co-articulatory effects of korean
language, The context-independent phoneme models are constructed using 51
phoneme-like units. Interpolation method is studied to combine the detailed
context-dependent phoneme models with robust context-independent phoneme
models and the concatenation of these interpolated phoneme models is applied to
the isolated word recognition. The experiment results show that the
context-dependent models lead to about 3.5% recognition improvement for the
context-independent phoneme models.