language: en
thumbnail: null
tags:
- audio-to-audio
- audio-source-separation
- Source Separation
- Speech Separation
- WHAM!
- REAL-M
- SepFormer
- Transformer
- pytorch
- speechbrain
license: apache-2.0
datasets:
- REAL-M
- WHAMR!
metrics:
- SI-SNRi
Neural SI-SNR Estimator
The Neural SI-SNR Estimator predicts the scale-invariant signal-to-noise ratio (SI-SNR) from the separated signals and the original mixture. The performance estimation is blind (i.e., no targets signals are needed). This model allows a performance estimation on real mixtures, where the targets are not available.
This repository provides the SI-SNR estimator model introduced for the REAL-M dataset.
Release | Test-Set (WHAMR!) average l1 error |
---|---|
18-10-21 | 1.7 dB |
Install SpeechBrain
First of all, please install SpeechBrain with the following command:
pip install speechbrain
Please notice that we encourage you to read our tutorials and learn more about SpeechBrain.
Minimal example for SI-SNR estimation
from speechbrain.pretrained import SepformerSeparation as separator
from speechbrain.pretrained.interfaces import SNREstimator as snrest
import torchaudio
# 1- Separate the mixture with a pretrained model (sepformer-whamr in this case)
model = separator.from_hparams(source="speechbrain/sepformer-whamr", savedir='pretrained_models/sepformer-whamr2')
est_sources = model.separate_file(path='speechbrain/sepformer-wsj02mix/test_mixture.wav')
# 2- Estimate the performance
snr_est_model = snrest.from_hparams(source="speechbrain/REAL-M-sisnr-estimator",savedir='pretrained_models/REAL-M-sisnr-estimator')
mix, fs = torchaudio.load('test_mixture.wav')
snrhat = snr_est_model.estimate_batch(mix, est_sources)
Inference on GPU
To perform inference on the GPU, add run_opts={"device":"cuda"}
when calling the from_hparams
method.
Training
The model was trained with SpeechBrain (fc2eabb7).
To train it from scratch follows these steps:
- Clone SpeechBrain:
git clone https://github.com/speechbrain/speechbrain/
- Install it:
cd speechbrain
pip install -r requirements.txt
pip install -e .
- Run Training:
cd recipes/REAL-M/sisnr-estimation
python train.py hparams/pool_sisnrestimator.yaml --data_folder /yourLibri2Mixpath --base_folder_dm /yourLibriSpeechpath --rir_path /yourpathforwhamrRIRs --dynamic_mixing True --use_whamr_train True --whamr_data_folder /yourpath/whamr --base_folder_dm_whamr /yourpath/wsj0-processed/si_tr_s
You can find our training results (models, logs, etc) here.
Limitations
The SpeechBrain team does not provide any warranty on the performance achieved by this model when used on other datasets.
Referencing SpeechBrain
@misc{speechbrain,
title={{SpeechBrain}: A General-Purpose Speech Toolkit},
author={Mirco Ravanelli and Titouan Parcollet and Peter Plantinga and Aku Rouhe and Samuele Cornell and Loren Lugosch and Cem Subakan and Nauman Dawalatabad and Abdelwahab Heba and Jianyuan Zhong and Ju-Chieh Chou and Sung-Lin Yeh and Szu-Wei Fu and Chien-Feng Liao and Elena Rastorgueva and François Grondin and William Aris and Hwidong Na and Yan Gao and Renato De Mori and Yoshua Bengio},
year={2021},
eprint={2106.04624},
archivePrefix={arXiv},
primaryClass={eess.AS},
note={arXiv:2106.04624}
}
Referencing SepFormer
@inproceedings{subakan2021attention,
title={Attention is All You Need in Speech Separation},
author={Cem Subakan and Mirco Ravanelli and Samuele Cornell and Mirko Bronzi and Jianyuan Zhong},
year={2021},
booktitle={ICASSP 2021}
}
About SpeechBrain
Website: https://speechbrain.github.io/
HuggingFace: https://huggingface.co/speechbrain/