Unit for Concatenation

Other methods using concatenative synthesis, especially those that do not modify the waveforms, use phonemes as their synthesis unit [2][3]. This is because fewer units are needed to synthesize any given new word, compared with the methods that adopt longer units. The method proposed in this paper uses syllables as the unit of synthesis though this leads to a compromise in the number of components. In our case, a syllable is a pair of a consonant (occasionally consonants or no consonant) and a vowel, which is normally recognized as a Kana character in Japanese.

1.

Synthesized words are used to provide information or to confirm information to users, therefore, clearness is a key issue. In recognizing Japanese syllables, the transition from the consonant to the vowel is said to be critical [4]. Thus this transition must be part of the synthesis unit.

2.

The words synthesized in this paper are mostly limited to proper nouns like personal names, station names, train names, city names, or town names. Because there is less need to consider prosodic variations of syllables than is true for sentence synthesis, the size of the syllabic database can be expected to be moderate.