Skip to main content
Figure 5 | EURASIP Journal on Advances in Signal Processing

Figure 5

From: Soft context clustering for F0 modeling in HMM-based speech synthesis

Figure 5

Normalized log-likelihood with respect to the number of leaves computed for each state of HMM. Blue and red curves are the learning curves of the hard and soft decision trees, respectively. In addition, solid curves illustrate the log-likelihood of the training set, and dashed curves are the log-likelihood computed for test data. MDL-based stop points are also shown through vertical dotted lines.

Back to article page