Kppwdfgu1/sbpn-snr-below-12db-20260722

Reliable SNR below 12 dB This gated review dataset contains 230 complete speech chunks with a reliable active-speech SNR estimate below 12 dB. The audio is cut from the original full recording—the same signal used for the SNR calculation—so the detected interference remains audible. The set covers the strong and severe noise bands. Each row includes a playable audio column, transcript text, speaker, timestamps, SNR, severity, and music metadata. Audio is mono 48 kHz, 96 kbps… See the full description on the dataset page: https://huggingface.co/datasets/Kppwdfgu1/sbpn-snr-below-12db-20260722.