VOICE INFORMATION PROCESSOR, AND ITS METHOD AND STORAGE MEDIUM

摘要

PROBLEM TO BE SOLVED: To generate a fragment dictionary which prevents deterioration in voice quality of voice synthesis. SOLUTION: A voice index retrieval part 401 retrieves a voice fragment which satisfies vocal sound environment from a voice database 210 and an HMM learning part 403 calculates HMM models for respective phonemes according to the retrieval result. A fragment recognition part 405 recognizes voice fragments according to the calculated HMM models for the respective phonemes and registers the fragment recognition result in the fragment dictionary when it is equal to the original voice fragment.