Reliability / Voice & Acoustics
Noise-Event Annotations
Annotators mark every audible noise event on each speaker's isolated track with a class and a start and end time, across 15 classes from mouth clicks and gulps to street traffic and thunderstorms. The same labels drive our own cleanup pipeline, and double as training data for noise detection, robust ASR and speech enhancement.
Specification
- Classes
- 15
- Granularity
- Event-level, start / end in seconds
- Scope
- Per speaker channel
- Tracks labelled
- 3.2K
Noise classes
- mouth click
- clicking sound
- background voices
- gulping
- microphone movement
- background noises
- tapping sound
- table touches
- chair movement
- body movement
- street vehicles
- outdoor noises
- finger clicking
- thunderstorm
- other
Sample pack
A curated set of takes from this dataset, with audio and annotation files in the delivery format.
More in Reliability / Voice & Acoustics
Voice & Speaker Profiles
Per-speaker accent, languages, age, gender, a self-described voice profile and recording setup, for voice design and balanced sampling.
49 voice actors
Raw / Clean Pairs
Takes as recorded and after noise reduction, with the noise-only segments that were profiled. Paired data for speech enhancement.
5.5K noise profiles