Experiments involving audio-based data like melspectrogram autoencoder, melspectrogram voice conversion.
Each folder store an entire field of experiments, so in each folder there is a specific README file with the experiment explanation.
- Melspectrogram Auto Encoder
- How to reconstruct melspectrograms? The bottleneck space can identify some voice pattern (emotions, gender, ...) ?
- Melspectrogram Conversion
- How to change an initital melspectrogram to another (neutral->happy)?
Lucas Hideki Ueda ([email protected])