Audio preprocessing

[anwielts]

Preprocess audio to train a ML model directly on the waveform data

[anwielts]

Open points:

• [x] Data augmentation directly on waveform
• [x] Truncate of fill waveform by repeating parts of the waveform
• [x] Oversampling under represented classes by applying more augmentation steps to Thema
• [ ] Balance sound types