Software Repository - Spatial Audio Scene Characterization (SASC): Automatic Localization of Front, Back, Up, and Down-Positioned Music Ensembles in Binaural Recordings

This repository consists of scripts and data useful to replicate the experiments described in the paper "Spatial Audio Scene Characterization (SASC): Automatic Localization of Front, Back, Up, and Down-Positioned Music En-sembles in Binaural Recordings".

Structure

The repository is organized as follows:

• models - the deep learning models that were trained in the study
• scripts - scripts used in the development of the deep learning algorithm: finding hyperparameters, model traning, evaluation, visualization, and statistical calculations
• data - partial results and detailed description of the data used in the study

Dependencies

Software dependencies:

• MATLAB - a development environment used to implement models
• VOICEBOX - a toolbox used to implement the binaural convolver
• SOFA - a file format for reading, saving, and describing spatially oriented data of acoustic systems.

Authors

Sławomir K. Zieliński ¹, Paweł Antoniuk ¹, and Hyunkook Lee ²

¹ Faculty of Computer Science, Białystok University of Technology, 15-351 Białystok, Poland; s.zielinski@pb.edu.pl (S.K.Z.); p.antoniuk6@student.pb.edu.pl (P.A.)

² Applied Psychoacoustics Laboratory (APL), University of Huddersfield, Huddersfield HD1 3DH, UK; H.Lee@hud.ac.uk (H.L.); D.S.Johnson2@hud.ac.uk (D.J.)

License

The content of this repository is licensed under the terms of the GNU General Public License v3.0 license. Please see the LICENSE file for more details.