TAU Spatial Sound Events 2019: Ambisonic and Microphone Array Evaluation Datasets

Two evaluation datasets containing four-channel audio recordings of stationary point sources, collected in Finland by Tampere University between December 2017 and June 2018. The datasets provide First-Order Ambisonic and tetrahedral microphone array recordings from identical scenes, with each sound event annotated with onset/offset times and direction-of-arrival coordinates. These datasets were created for the DCASE 2019 Sound Event Localization and Detection Task.

Use Cases

• Benchmarking sound event localization models based on annotated direction-of-arrival coordinates.
• Comparing audio encoding formats for spatial tasks using the parallel Ambisonic and microphone array recordings.
• Training machine learning models for acoustic scene analysis based on multi-channel recordings of labeled sound events.

Strengths

• Provides two synchronized four-channel audio formats (First-Order Ambisonic and tetrahedral microphone array) from the same scenes, enabling format comparison.
• Data collection received funding from the European Research Council grant agreement 637422 EVERYSOUND, indicating a research-grade origin.
• Recordings are annotated with temporal onset/offset and spatial direction-of-arrival coordinates (azimuth and elevation).

Limitations

• Row count and total dataset size are unknown, which may limit suitability assessment for large-scale training.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Tampere University

Collection Method

Recordings collected using a microphone array.

Time Range

12/2017 - 06/2018

Geography

Finland