BorisovMaksim/denoising
This is a repo that implements web interface for DEMUCS model proposed in Real Time Speech Enhancement in the Waveform Domain. The model was trained from scratch in Pytorch. The proposed model is based on an encoder-decoder architecture with skip-connections. It is optimized on both time and frequency domains, using multiple loss functions. You can record your voice in noisy conditions and get denoised version using DEMUCS model. There is also Spectral Gating denoiser as baseline.
<center><img src="https://th.bing.com/th/id/OIP.JuP5GDiGXTGkVR-x7Ov0bgHaF8?w=233&h=186&c=7&r=0&o=5&dpr=2&pid=1.7" > </center>
Running
Without docker:
<pre><code>pip install -r requirements.txt python app.py</code></pre>
Using docker: <pre><code>docker build . --tag python-docker docker run -p 7860:7860 -e GRADIOSERVERNAME=0.0.0.0 -it python-docker:latest</code></pre>
Data
In the scope of this project Valentini dataset in used. It is clean and noisy parallel speech database. The database was designed to train and test speech enhancement methods that operate at 48kHz. There are 56 speakers and ~10 gb of speech data.
For model improvement it is possible to use a bigger training set from DNS challenge.
Training
The training process in impemented in Pytorch. The data is (noisy speech, clean speech) pairs that are loaded as 2 second samples, randomly cutted from audio and padded if necessary. Model is optimized using SGD. In terms of loss functions, the L1 loss and MultiResolutionSTFTLoss are used. MultiResolutionSTFTLoss is the sum of STFT loss over different window sizes, hop sizes and fft sizes.
$$L{STFT}= L{sc} + L_{mag}$$
$$L{sc}= \frac{|| |STFT(\tilde{x})| - |STFT(x)| ||{F}^{1}}{|STFT(x)|}$$
$$L{mag} = \frac{1}{T}|| log|STFT(\tilde{x})| - log|STFT(x)| ||{F}^{1}$$
where T is the time points in the waveform.
Metrics
The PESQ metric is used for estimating overall speech quality after denoising and STOI is used for estimating speech intelligibility after denoising. Intelligibility measure is highly correlated with the intelligibility of degraded speech signals
Experiments
For tracking experiments local server of Weights & Biases is used. To manage configs for different experiments hydra is used. It allows an easy way to track configs and override paramaters.