TAU-SRIR DB: Spatial Room Impulse Responses from Multiple Rooms

The TAU Spatial Room Impulse Response Database contains 32-channel spatial room impulse responses captured at Tampere University, Finland. It includes responses for multiple source positions and distances per room, along with ambient noise recordings, collected between December 2017 and January 2020. The dataset is intended for emulating spatial multichannel recordings to evaluate and train audio processing algorithms.

Use Cases

• Training multichannel speech enhancement models based on high-resolution 32-channel impulse responses.
• Evaluating sound event localization and detection methods based on emulated moving source scenarios from densely spaced measurement trajectories.
• Testing acoustic scene analysis algorithms based on data captured in multiple rooms with varying acoustic conditions.
• Generating binaural or Ambisonics audio mixtures for machine listening tasks based on derived application-specific formats.

Strengths

• Contains high-resolution 32-channel spatial room impulse responses, from which multiple application-specific formats can be derived.
• Includes multiple source distances, azimuths, and elevations per room, allowing emulation of complex multi-source configurations.
• Data was captured in multiple rooms, enabling evaluation and training for generalization across different acoustic conditions.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The database is described as a work-in-progress, with future versions planned.

Provenance

Source

Tampere University

Collection Method

Spatial room impulse responses were captured by staff for a fixed receiver and multiple source positions per room.

Time Range

Data collected between 12/2017 - 06/2018 and 11/2019 - 01/2020.

Geography

Various spaces at Tampere University, Finland.