We release the Expressive Anechoic Recordings of Speech (EARS) dataset.
If you use the dataset or any derivative of it, please cite our research paper:
@inproceedings{richter2024ears,
title={{EARS}: An Anechoic Fullband Speech Dataset Benchmarked for Speech Enhancement and Dereverberation},
author={Richter, Julius and Wu, Yi-Chiao and Krenn, Steven and Welker, Simon and Lay, Bunlong and Watanabe, Shinjii and Richard, Alexander and Gerkmann, Timo},
booktitle={ISCA Interspeech},
year={2024}
}
The dataset is available for download from https://github.com/facebookresearch/ears_dataset.
The EARS-WHAM and EARS-Reverb datasets are available for download from https://github.com/sp-uhh/ears_benchmark.
Audio examples can be found at our project page.