ABCS is an Air- and Bone-Conducted synchronized Speech corpus.
The corpus is composed of 47,182 AC and BC synchronized utterances from 100 speakers, which amounts to 42 hours of speech with each speaker contributing approximately 25 minutes. The duration of each utterance varies from 1 to 5 seconds.
The corpus was collected in an anechoic chamber, where the speakers wore the headset and read the text transcription. The anechoic chamber was built in accordance with the ISO 3745. The size of the anechoic chamber is 11.8×4.2×3.8 m.
If you used ABCS corpus and you want to cite us, use this :
@ARTICLE{Wang2022ABCS,
author={Wang, Mou and Chen, Junqi and Zhang, Xiao-Lei and Rahardja, Susanto},
journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing},
title={End-to-end Multi-modal Speech Recognition On An Air and Bone Conducted Speech Corpus},
year={2022},
volume={},
number={},
pages={1-12},
doi={10.1109/TASLP.2022.3224305}
}