At the end of last month Marc Green and Prof. Damian Murphy published the journal paper “EigenScape: A Database of Spatial Acoustic Scene Recordings”, which introduces the EigenScape database.
Marc recorded EigenScape as part of his PhD research looking into machine learning for environmental sound. The database features acoustic scenes recorded at locations across the North of England using the mh Acoustics Eigenmike, and is the largest database of spatially-recorded soundscapes publicly available.
Marc was able to show that the spatial properties of acoustic scenes can be utilised to train computers to detect the type of location in which the recording was made – the first piece of research to use spatial features to do this. It is the hope that the database will stimulate further research into machine listening in spatial audio.
EigenScape is available for download at http://doi.org/10.5281/zenodo.1012809
Let us know if you find it useful!