Libri-AudioEvent is a synthesized noisy-speech dataset containing matched noisy speech, clean speech, and noise signals.
All clips are sampled at 16 kHz with 10s duration.
Dataset Structure
The repository contains 4 splits:
training
validation
test_mix
test_iso
Each split file contains:
metadata.csv
noisy/
clean/
noise/
Scale of data:
Training
Validation
Test-Mix
Test-Iso
10,800
500
200
200
Extract Dataset… See the full description on the dataset page: https://huggingface.co/datasets/Ediethia/Libri-AudioEvent.
