Figure 3From: Soft context clustering for F0 modeling in HMM-based speech synthesisSamples and decision tree structures. (a) Data samples used to train the decision trees. The red dots and blue line are training samples and objective function, respectively. (b) Trained hard decision tree structure. (c) Trained soft decision tree structure. (d) Predicted samples using the hard decision tree. (e) Predicted samples using the soft decision tree.Back to article page