| 摘要 |
PROBLEM TO BE SOLVED: To increase the compression rate of the compression-encoding of voices and to resolve the discontinuity of the pitch while the phoneme pieces, which are compressed-encoded, are edited and synthesized voices are outputted. SOLUTION: The pitches of inputted voices are analyzed.extracted in an analysis interval unit. The extracted pitches are then made into a model by the addition of the accents, the phrases and the patterns of phonemes stored in a pitch pattern ROM 33. Moreover, a pitch error extractor 34 extracts the error between the pitch of actual voice and the patterned pitch and generates compressed and coded voice code data by transmitting each pattern and its error. The decoder generates.decodes pitches by adding the accents.the phrases.each pattern of phonemes and the error between the pattered pitch and the pitch of the actual voice. |
