Improving quality in a speech synthesizer based on the MBROLA algorithm.
Borja EtxebarriaInmaculada HernáezI. MadariagaEva NavasJ. C. RodríguezR. GándaraPublished in: EUROSPEECH (1999)
Keyphrases
- preprocessing
- improved algorithm
- experimental evaluation
- learning algorithm
- optimal solution
- times faster
- tree structure
- detection algorithm
- high accuracy
- computational cost
- text to speech
- convex hull
- simulated annealing
- computational complexity
- optimization algorithm
- data sets
- recognition algorithm
- clustering method
- dynamic programming
- expectation maximization
- input data
- linear programming
- np hard
- computationally efficient
- theoretical analysis
- segmentation algorithm
- k means
- significant improvement
- search space
- classification algorithm
- hidden markov models
- search algorithm
- worst case
- quality measures
- similarity measure
- estimation algorithm