Ryuichi Yamamoto

migrate from my C++ code

https://github.com/sarulab-speech/tdmelodic_openjtalk

https://github.com/r9y9/pyopenjtalk/blob/d28b9108b29b857db8efe153ac483efe38afc2a4/pyproject.toml#L16

https://r9y9.github.io/nnmnkwii/latest/nnmnkwii_gallery/notebooks/tts/01-DNN-based%20statistical%20speech%20synthesis%20(en).html ```py subplot(4,1,3) lf0 = y[:,mgc_start_idx] vuv = y[:,vuv_start_idx] plot(lf0, linewidth=2, label="Continuous log-f0") plot(vuv, linewidth=2, label="Voiced/unvoiced flag") legend(prop={"size": 14}, loc="upper right") ``` ```py lf0 = y[:,mgc_start_idx] ``` should be ```py...

Adaptive mel-generalized cesptrum analysis - [ ] acep - [ ] agcep - [ ] amcep Mel-generalized cepstrum analysis - [ ] uels - [ ] fftcep LPC, LSP and...

To avoid waf-related errors during waf configurations

Under 50% for now. Over 80% (ideally, 100%) is better.

- [ ] Global variance enhancement - [ ] Modulation spectrum enhancement