A preselection method based on cost degradation from the optimal sequence for concatenative speech synthesis.
Nobuyuki NishizawaHisashi KawaiPublished in: INTERSPEECH (2007)
Keyphrases
- significant improvement
- dynamic programming
- pairwise
- high accuracy
- detection method
- speech synthesis
- preprocessing
- cost function
- prediction error
- input data
- expected cost
- exhaustive search
- closed form
- classification method
- synthetic data
- clustering method
- computer vision
- experimental evaluation
- pattern recognition
- objective function