System overview

Audio samples

RAW (Target)

bdl

rms

clb

slt

SD WaveNet Vocoder

bdl

rms

clb

slt

SI-CLOSE WaveNet Vocoder

bdl

rms

clb

slt

SC WaveNet Vocoder

bdl

rms

clb

slt

SI-OPEN WaveNet Vocoder

bdl

rms

clb

slt

You can download more examples from here

Subjective Evaluation Result

Experimental condition

• #evaluation speakers=4
• #evaluation utts per speaker=40
• #subjects=9
• #evaluation utts per subject=120

Experimental Result

The effect of the amount of training data

References

@article{hayashi2018sp,
  title={複数話者WaveNetボコーダに関する調査}.
  author={林知樹 and 小林和弘 and 玉森聡 and 武田一哉 and 戸田智基},
  journal={電子情報通信学会技術研究報告},
  year={2018}
}
@article{hayashi2017multi,
  title={An Investigation of Multi-Speaker Training for WaveNet Vocoder},
  author={Hayashi, Tomoki and Tamamori, Akira and Kobayashi, Kazuhiro and Takeda, Kazuya and Toda, Tomoki},
  journal={Proc. ASRU 2017},
  year={2017}
}
@inproceedings{tamamori2017speaker,
  title={Speaker-dependent WaveNet vocoder},
  author={Tamamori, Akira and Hayashi, Tomoki and Kobayashi, Kazuhiro and Takeda, Kazuya and Toda, Tomoki},
  booktitle={Proceedings of Interspeech},
  pages={1118--1122},
  year={2017}
}

Contact

Tomoki Hayashi @ Nagoya University
E-mail: hayashi.tomoki at g.sp.m.is.nagoya-u.ac.jp