ARVERIR: The Sound of the Violin's Home — A Higher-Order Room Impulse Response Dataset of the Arvedi Auditorium in Cremona

Official repository of the ARVERIR dataset and the related article.

ARVERIR is a dataset of spatial room impulse responses (RIRs) measured in the Giovanni Arvedi Auditorium in Cremona, Italy, a contemporary concert hall located within the Museo del Violino complex and dedicated to the performance and recording of string instruments. Measurements were acquired using higher-order microphones (Voyage Audio Spatial Mic Dante), 8-capsule spherical arrays placed at 65 receiver positions across the audience area, while 5 sources were deployed on stage: an omnidirectional reference (Brüel & Kjær OmniPower Type 4292-L) at the geometric center of the stage, and 4 directional loudspeakers (Genelec 8020C) arranged at the canonical positions of a string quartet or small ensemble.

• Read the paper 📄 — 2026 International Workshop on Acoustic Signal Enhancement (IWAENC), link available soon
• Download the dataset ⬇️

Dataset at a glance

• concert-hall measurements in the Arvedi Auditorium, inside the Museo del Violino complex in Cremona
• 5 sources on stage, identified as S0 (omnidirectional reference) and S1–S4 (directional loudspeakers)
• 65 higher-order microphone positions distributed over 5 rows of the audience area
• 325 source–receiver spatial RIRs
• 8-channel RIRs at 48 kHz, 2 s long, stored as 32 bit wav files
• calibrated source, array and capsule positions provided as csv files

RIRs were obtained by deconvolving logarithmic sine sweeps ranging from 50 Hz to 22 kHz, each lasting 10 s and sampled at fs = 48 kHz. They are provided as 8-channel wav files saved at 32 bit per sample, with one channel per capsule of the higher-order microphone, and truncated to a duration of 2 s. The capsule-domain representation was chosen to preserve the raw spatial information and leave users free to apply their preferred Ambisonic encoding, calibration, or array processing pipeline. Together with RIRs, we also release acoustically calibrated positions of each source, of each array center and of each microphone capsule.

Dataset content

The dataset is distributed on Zenodo as a single archive (904 MB) that expands to the following structure:

arvedi_auditorium_dataset/
├── pos_sources.csv
├── pos_receivers.csv
└── rirs/
    ├── rir-S0-A101.wav
    ├── rir-S0-A102.wav
    ├── ...
    └── rir-S4-A512.wav

Sources

Five sources were placed on stage. S0 is an omnidirectional reference source, whose position defines the origin of the coordinate system; S1–S4 are directional loudspeakers placed at the canonical positions of a string quartet, with S1 and S2 corresponding to the first and second violin, S3 to the viola and S4 to the cello. Coordinates are expressed in meters.

Source	Model	x [m]	y [m]	z [m]
S0	B&K OmniPower Type 4292-L	0.00	0.00	1.45
S1	Genelec 8020C	−0.05	−1.99	1.18
S2	Genelec 8020C	−1.17	−1.09	1.18
S3	Genelec 8020C	−1.32	0.91	1.18
S4	Genelec 8020C	−0.46	2.20	1.18

Receivers

Measurements cover sector A of the audience area, over five rows of 10, 12, 14, 16 and 12 seats respectively, for a total of 64 receiver positions. For each position the microphone was placed at the location of the corresponding seat, at a height of 1.16 m above the floor of the row. One additional off-grid measurement was performed between seats 9 and 10 of row 5, intended to facilitate the validation of sound field reconstruction and interpolation algorithms, bringing the total to 65 receiver positions.

Each receiver is a Voyage Audio Spatial Mic, a higher-order microphone featuring 8 pre-polarized cardioid condenser capsules arranged on a rigid spherical baffle of about 2.1 cm of radius, allowing higher-order Ambisonic encoding up to the 2nd order.

Naming convention

File names follow a structured, hyphen-separated convention in which each token conveys specific information about the file content and the associated measurement configuration.

The dataset adopts the following file name pattern:

rir-S<source>-A<row><seat>.wav

with the following meaning:

• rir denotes a room impulse response measurement
• S<source> identifies the source, with <source> ranging from 0 to 4
• A<row><seat> identifies a receiver position, where:
  • A denotes sector A of the audience area
  • <row> is the row index, ranging from 1 to 5
  • <seat> is the zero-padded seat number within the corresponding row

The additional off-grid measurement, taken between seats 9 and 10 of row 5, is named rir-S<source>-A509_A510.wav.

Position files follow the conventions pos_sources.csv for source positions and pos_receivers.csv for receiver positions.

Examples:

• rir-S0-A101.wav denotes the RIR measured from the omnidirectional source S0 to the receiver at row 1, seat 01
• rir-S3-A412.wav denotes the RIR measured from source S3 to the receiver at row 4, seat 12

Position files

Both position files are semicolon-separated, with one header row and one row per source or receiver.

pos_sources.csv:

Column	Description
source	Source identifier, S0 to S4
x,y,z	Coordinates of the source [m]

pos_receivers.csv:

Column	Description
mic	Receiver identifier, e.g. A305
C_x,C_y,C_z	Coordinates of the center of the array [m]
x1_x,x1_y,x1_z	Coordinates of capsule 1, i.e. of channel 1 [m]
...	...
x8_x,x8_y,x8_z	Coordinates of capsule 8, i.e. of channel 8 [m]

Sample code

The repository contains two sample scripts for reading and visualizing both RIRs and microphone/source positions:

• arverir.py: Python example script (requires numpy, scipy and matplotlib)
• arverir.m: MATLAB example script

Each script reads one RIR and the position metadata, prints the properties of the measurement, and produces two figures: the impulse responses of the 8 capsules around the direct sound, and the measurement geometry seen from above.

The dataset should be downloaded separately from Zenodo. By default, both scripts assume that the dataset is available in arvedi_auditorium_dataset; this path can be adjusted directly in the scripts if needed, together with the source, receiver and channel to inspect.

Citation

If you use this dataset, please cite:

F. Miotello et al., "The Sound of the Violin's Home: A Higher-Order Room Impulse Response Dataset of the Arvedi Auditorium in Cremona," 2026 International Workshop on Acoustic Signal Enhancement (IWAENC), 2026.

BibTeX:

@inproceedings{miotello2026arverir,
  title={The Sound of the Violin's Home: A Higher-Order Room Impulse Response Dataset of the Arvedi Auditorium in Cremona},
  author={Miotello, Federico and Greco, Gioele and Ostan, Paolo and Del Gaudio, Francesca and Comanducci, Luca and Malvermi, Raffaele and Pezzoli, Mirco and Antonacci, Fabio},
  booktitle={2026 International Workshop on Acoustic Signal Enhancement (IWAENC)},
  year={2026},
  organization={IEEE}
}