المستخلص: |
To approch a human voice, some modifications on speech signal produced by ITS systems are necessary. As we know, the relevant parameters of the vocal signal are: the fundamental frequency, energy and duration. In our work, we have studied the duration of segmental phonemes. In fact, we tried to reduce the different imperfections which depend on duration of voiced consonants having unvoiced oppositions of arabic language. For that, we measured first different coefficients of variation (position and context), and after, we constructed our model of correction. We tested this model to modify the duration of some phonemes (voiced) produced by Text-to-Speech system.
|