schism-audio/rirs-noises

RIRS NOISES

Quick Start

from datasets import load_dataset

# Stream to avoid downloading the entire dataset
ds = load_dataset("schism-audio/rirs-noises", streaming=True)

# Or download locally
ds = load_dataset("schism-audio/rirs-noises")

Dataset Description

RIRS NOISES is the real/isotropic RIR and point-source noise subset from OpenSLR SLR28. This Hugging Face repo contains 1,260 WAV files: 417 files under real_rirs_isotropic_noises/ and 843 files under pointsource_noises/. All audio is 16 kHz WAV.

This subset is useful for audio data augmentation: convolving clean audio with real RIRs to simulate reverberant environments and adding noise at controlled SNRs. For the full 61,260-file OpenSLR SLR28 mirror, including simulated RIRs, see schism-audio/openslr-rirs.

Dataset Structure

Data Fields

Field	Type	Description
audio	Audio	WAV file at 16 kHz, 16-bit
label	ClassLabel	Top-level source directory: pointsource_noises or real_rirs_isotropic_noises

Data Splits

This dataset has no predefined splits. All files are in the default train split.

Split	Files
train	1,260

File counts by directory:

Directory	Files
pointsource_noises/	843
real_rirs_isotropic_noises/	417

Usage Examples

Load RIRs for augmentation

from datasets import load_dataset

ds = load_dataset("schism-audio/rirs-noises")

# Filter for real RIRs only
rirs = ds["train"].filter(
    lambda x: ds["train"].features["label"].int2str(x["label"]) == "real_rirs_isotropic_noises"
)
print(f"Real RIRs: {len(rirs)}")

Convolve audio with an RIR

import numpy as np
from scipy.signal import fftconvolve
from datasets import load_dataset

ds = load_dataset("schism-audio/rirs-noises")
rir = ds["train"][0]["audio"]["array"]

# Normalize the RIR
rir = rir / np.max(np.abs(rir))

# Convolve with your dry audio signal
# reverberant = fftconvolve(dry_audio, rir, mode="full")

Dataset Creation

Source Data

The dataset was compiled from multiple publicly available acoustic databases:

• Real/isotropic RIR and noise files: 417 WAV files under real_rirs_isotropic_noises/.
• Point-source noises: 843 WAV files under pointsource_noises/.

Annotations

No manual annotations are included. File organization and naming encode the source database and recording type.

Known Limitations

• Speech-centric design: The RIRs and noise profiles were originally curated for speech processing. Drum and music applications may require additional filtering or selection.
• 16 kHz only: The fixed 16 kHz sample rate may be limiting for music applications that require higher fidelity.
• Subset only: This repo does not include the 60,000 simulated RIR files from OpenSLR SLR28. Use schism-audio/openslr-rirs for the complete mirror.

Related Datasets

This dataset is part of the Drum Audio Datasets collection by schism-audio. Related datasets:

• schism-audio/dechorate — Calibrated multichannel RIRs with echo annotations and 3D positions

Citation

@misc{rirs_noises,
  title     = {A database of room impulse responses and noise recordings},
  author    = {Ko, Tom and Peddinti, Vijayaditya and Povey, Daniel and Seltzer, Michael L. and Khudanpur, Sanjeev},
  year      = {2017},
  url       = {https://openslr.org/28/}
}

License

This dataset is released under the Apache License 2.0.

You are free to use, modify, and distribute this dataset for any purpose, including commercial use, subject to the terms of the Apache 2.0 license.