Table 4 The usages of the subsets for each speech corpus and their statistics
From: A parametric prosody coding approach for Mandarin speech using a hierarchical prosodic model
Corpus | Subsets | Usages | Spk# | Utt# | Syl# | Hours | Remark |
|---|---|---|---|---|---|---|---|
Treebank | TrainTB | Training of the HPM, the AM for forced-alignment and the models for HMM-based speech synthesizer | 1 | 376 | 51,868 | 3.9 | Â |
TestTB | Evaluation of prosody coding | 1 | 44 | 3898 | 0.3 | Â | |
TCC300 | TrainTC1 | Training of the AM for forced-alignment | 274 | 8036 | 300,728 | 23.9 | Include all set A and 90% of set B |
TrainTC2 | Training of the HPM | 164 | 962 | 106,955 | 8.3 | Subset of TrainTC1 | |
TestTC | Evaluation of prosody coding and adaptation of HMM model for speech sythesis | 19 | 226 | 26,357 | 2.4 | Selected from Set B of TCC300 |