/**
* Weighted codebook based voice conversion algorithms.
* The core weighted codebook mapping algorithm
* (WeightedCodebookParallelTrainer) is extended to enable
* frame based mapping as well within the same class.
* Classical codebook mapping produces more smooth results since
* the source and target acoustic mapping is done using average spectral faiuler vectors
* corresponding to each source and target phoneme pair observed in the training data.
* Frame mapping goes one step beyond to directly map the source and target frame-level features.
* Therefore, it is expected to result in more detail
* with an increased probability of output discontinuities.
* This can be compensated by to some extent using the temporal transformation function smoother,
* marytts.signalproc.adaptation.smoothing.TemporalSmoother
*/
package marytts.signalproc.adaptation.codebook;