NoisySpeech-CCMix-Industrial v1 (mini)
收藏Zenodo2025-09-18 更新2026-05-26 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.17156195
下载链接
链接失效反馈官方服务：
资源简介：
NoisySpeech-CCMix-Industrial v1 (mini) is a compact synthetic noisy-speech benchmark for denoising, robust ASR, and source separation. Clean speech is sourced from LibriSpeech test-clean (CC-BY 4.0). We mix each utterance with algorithmic noises—white, pink, band-limited pink (“bandpink”), machinery-like hum (AM/FM tonal), sparse impacts, and siren—at continuous SNRs in [-8, +15] dB. Some mixtures include multiple noises; exact noise types and mixture weights are recorded. Audio is 16 kHz mono, FLAC. The dataset includes deterministic train/val/test splits and a comprehensive metadata.csv with requested/achieved SNR, noise types, weights, durations, and RMS values.
Intended use: benchmarking speech enhancement and robust ASR under industrial & multi-noise conditions. This “mini” release prioritizes quick download and reproducibility; a larger “full” release is planned.
Attribution & license: This dataset is CC-BY 4.0 due to the LibriSpeech upstream license. Please retain ATTRIBUTION.txt and cite both this dataset’s DOI and the LibriSpeech paper/source.
提供机构：
Zenodo
创建时间：
2025-09-18
