Shorter synthesis units lower concatenation cost but syllables are the optimal balance
Evidence
When designing an implementation of unit-selection synthesis, the size of the units may be set to be shorter or longer. If they are made shorter, more joining points will be available to select the next unit. Since there are more possibilities to select the joining point, there will be a smaller concatenation cost, which leads to better quality. However, too short a length for the units may also cause problems. Kishore and Black (2003) tested the optimal length for the Hindi language, and the best result was obtained with the unit length equalling that of syllables. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 2248
Categorization
Category: Causal Relationships
Related
- Unit-selection synthesis achieves naturalness at the cost of huge databases and inflexibility (similarity 0.83)
- Speech synthesis selects units by minimizing target and concatenation costs. (similarity 0.82)
Broader Connections
- Speech is a continuous signal flow, not discrete concatenated units
#complicates(0.80) — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 565 - Syllables have stronger audiovisual coherence than sentences, predicting greater binding.
#corroborates(0.75) — Dijk, Pim van, Physiology, Psychoacoustics and…, 5137