Skip to content

Latest commit

 

History

History
27 lines (25 loc) · 1.16 KB

README.md

File metadata and controls

27 lines (25 loc) · 1.16 KB
Raw

Input files structure

Audio files belonging to each class should be placed together in a folder named as the label for that class. No other labeling is necessary. Example:

disvoice/
├── disvoice_character.py
├── disvoice_features.md
├── TONE_AGITATED
│   ├── agitated_sample_1.wav
│   ├── agitated_sample_2.wav
│   |── agitated_sample_3.wav
|   ...
└── TONE_RELIEVED
    ├── relieved_sample_1.wav
    ├── relieved_sample_2.wav
    |── relieved_sample_3.wav
    ...

Audio characteristics Extraction

  • Specify the data folder names through arguments. From example above:
    $ python disvoice_character.py --categories TONE_AGITATED TONE_RELIEVED
  • 3 types of features are statically extracted from each audio file: articulation, phonation, prosody
  • Output file stores shared audio labels & 3 feature matrices collected independently from each feature type

DisVoice-Specific Dependencies

  • Praat: obtain barren server executable; create symbolic link praat -> praat-barren in a PATH visible location
  • Python=3.10; NumPy=1.22.4; SciPy=1.11.4