摘要 |
PURPOSE:To easily obtain a natural synthesized speech by determining a mora interval by using a consonant between two moras as a parameter, and adjusting the phoneme continuance of a document to be voiced with the mora interval. CONSTITUTION:The relation between the time length between the centroid of energy of vowel parts and the consonant between the vowels is stored as the rhythm rule of a phoneme continuance length generation part 2. The phoneme continuance length generation part 2 integrates the times of energy of speech waveforms of two adjacent moras (units of phoneme corresponding to monosyllable unit) in a phoneme symbol series generated by a text analysis part 1. Then the mora interval is found from the time length between vowel part energy integration. The mora interval is determined by a rhythm rule and the phoneme continuance is adjusted with the mora interval and outputted. The power pattern of a speech is determined by a sound source amplitude pattern generation part 3 according to the phoneme continuance length and a continuous point pitch pattern is generated by a pitch pattern generation part 4. |