mcw519 / PureSound

Make the sound you hear pure and clean by deep learning.
7 stars 0 forks source link
deep-learning end-to-end pytorch speech-enhancement speech-separation target-speaker-extraction

PureSound

We hope this repo can help you to listen pure clean voice/sound.

Install & Test

git clone <project-url>
cd puresound && sh build_puresound.sh

Repo struct & Recipes

PureSound
├─puresound
│  ├─nnet: DNN realted models and loss function.
│  │  ├─lobe
│  │  └─loss
│  ├─src: Audio I/O and others processing.
│  ├─streaming: DNN models streaming structure.
│  └─task: PyTorch trainer and Dataset.
├─egs
│  ├─ns: Example of Noise Supression
│  └─tse: Example of Target Speech Extraction
│      ├─demo: Simple demo App for real-time recording and extract your voice