TAU Urban Acoustic Scenes 2022 Mobile: 64 Hours of 1-Second Audio Segments

64 hours of audio are organized into 1-second segments across 10 distinct urban acoustic scenes. The development dataset was created by Toni Heittola of Tampere University for the DCASE 2022 challenge. It focuses on audio captured with mobile devices in environments like airports, metro stations, and pedestrian streets.

Use Cases

• Train acoustic scene classification models based on the 10 labeled urban environments.
• Benchmark audio event detection algorithms on short, 1-second audio segments.
• Study urban soundscape characteristics using data from mobile device recordings.

Strengths

• Contains 64 hours of audio data, providing a substantial corpus for model training.
• Focuses on 10 specific urban acoustic scenes, enabling targeted classification tasks.
• Audio is segmented into consistent 1-second clips, which may simplify preprocessing.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• The specific recording locations and device details are not described, which may limit reproducibility.

Provenance

Source

Tampere University

Collection Method

Likely consists of audio recordings collected with mobile devices in various urban settings.