Percetual Sound Browser is a demo prototype of a sound browser which enables you browsing sounds by their high-level acoustic/perceptual qualities such as brightness, roughness, hardness and depth. All these features are computed from the original audio content itself, locally. This was developed as a probe system in further investigating the scope of integrating high-level semantic features into audio production environments.
For the moment, this patch only uses a small customised dataset of 700 sounds (7 classes) from Freesound. But, you can integrate any freesound collections by using the attached feature extraction python scripts in the instructions.
This project uses early versions of the timbral sound models developed by Andy Pearce (IoSR, University of Surrey) for off-line feature extraction. It was developed by Albin Andrew Correya at the Music Technology Group of Universitat Pompeu Fabra as part of the AudioCommons project.