The library was created by identifying and isolating 10 s audio segments thatrepresent discrete sound categories from the following sources: BBC Complete Sound Effects Library(bbcsfx.acropolis.org.uk) and freesound.org. Particular care has been taken in the selection of thesegments in order to keep the sound samples clean of background noise and prevent overlapping ofthe sound classes.
The library is organized in the following 10 classes: airplanes, alarms, applause,birds, dogs, motorcycles, rain, rivers, sea waves, and thunders.
Our intention was to include classes ofsounds—encountered either in indoor or outdoor soundscapes—that are extensively used in similarrecognition schemes. Each class is equally represented in the database by 18 audio files with greatvariations between them, which reflect real life situations. All the recordings are uncompressed monoaudio files in WAV format, with a sampling rate of 44.1 kHz and 16 bit analysis.
Publications
[1] Bountourakis, V., Vrysis, L., & Papanikolaou, G. (2015). Machine learning algorithms for environmental sound recognition: Towards soundscape semantics. In Proceedings of the Audio Mostly 2015 on Interaction With Sound (pp. 1-7).
[2] Bountourakis, V., Vrysis, L., Konstantoudakis, K., & Vryzas, N. (2019, June). An Enhanced Temporal Feature Integration Method for Environmental Sound Recognition. In Acoustics (Vol. 1, No. 2, pp. 410-422). Multidisciplinary Digital Publishing Institute.
Licence
If you wish to use the dataset for academic purposes, please cite [1] and [2]
Downloads