This project is associated with the recently-released AIHSHELL-4 dataset for speech enhancement, separation, recognition and speaker diarization in conference scenario. The project, served as baseline, is divided into five parts, named data_preparation, front_end, asr and sd. The Speaker Independent (SI) task only evaluates the ability of front end (FE) and ASR models, while the Speaker Dependent (SD) task evaluates the joint ability of speaker diarization, front end and ASR models. The goal of this project is to simplify the training and evaluation procedure and make it easy and flexible for researchers to carry out experiments and verify neural network based methods.
git clone https://github.com/felixfuyihui/AISHELL-4.git
pip install -r requirements.txt
If you use this challenge dataset and baseline system in a publication, please cite the following paper:
@article{fu2021aishell,
title={AISHELL-4: An Open Source Dataset for Speech Enhancement, Separation, Recognition and Speaker Diarization in Conference Scenario},
author={Fu, Yihui and Cheng, Luyao and Lv, Shubo and Jv, Yukai and Kong, Yuxiang and Chen, Zhuo and Hu, Yanxin and Xie, Lei and Wu, Jian and Bu, Hui and Xin, Xu and Jun, Du and Jingdong Chen},
year={2021},
conference={Interspeech2021, Brno, Czech Republic, Aug 30 - Sept 3, 2021}
}
The paper is available at https://arxiv.org/abs/2104.03603
Dataset is available at http://www.openslr.org/111/ and http://www.aishelltech.com/aishell_4