3. New Positional Features Replacing Prosodic Factors

The conventional methods [2][3] consider prosodic factors of the synthesis units like fundamental frequency (f₀), power, and duration. Unfortunately, it is extremely complicated to accurately analyze and appropriately estimate each factor. When we synthesize isolated words, however, we can simplify the procedure while realizing a voice with adequate quality that ensures customer satisfaction. We introduce the new concept of ``positional features''. These are used to characterize the prosodic features instead of the parameters mentioned above. Positional features are the position of a component syllable within a given word and the mora length of the word. Thus, a syllable component is expressed as $\{Sy(P,N)_{m,M},{\rm index~to~waveform}\}$, where Sy represents a syllable, P represents the prior phoneme, N represents the posterior phoneme, m represents the mora position of Sy in the word, and M represents the mora length of a word. We will show how these features can well replace the conventional prosodic features.