摘要 |
PURPOSE: A unit phoneme coding method is provided to preserve phase information of phonemes when coding the phonemes, and to restore the preserved phase information when decoding the phonemes, in order to prevent phase discontinuity, thereby reducing size of a unit sound database. CONSTITUTION: A system extracts envelope information from each phoneme stored in an original unit sound database(30). The system extracts phase information of excitation signals included in the phonemes(31). The system stores the envelope information and the phase information in a compressed unit sound database by phoneme(32). When extracting the envelope information from each phoneme stored in the original unit sound database, the system uses a sub-band CELP(Code Excited Linear Predictive) coding method in order to compress high-quality speech data. |