Soundscape Audio Signal Classification and Segmentation UsingLIstenrs Perception of Background and Foreground Sound 

Miles Thorogood, Jianyu Fan, Philippe Pasquier

Abstract

Classification and segmentation are important but time-consuming tasks when using sound-scape recordings in sound design and research. Background and foreground are criteria when segmenting sound files according to a signal’s perceptual attributes. We establish the back- ground and foreground classification task within a musicological and soundscape context, and present a method for the automatic segmentation of soundscape recordings based on this task. We present a soundscape corpus with ground truth data obtained from a human perception study. An analysis of the corpus shows participants have a high level of agreement on the category assigned to background samples (92.5%), foreground samples (80.8%), and background with foreground samples (75.3%). We verify the corpus by training a Support Vector Ma- chines classifier. An analysis of the classifier demonstrates a similarly high degree of certainty for background 96.7%, foreground 80%, and background with foreground 86.7%. Further, we report an experiment evaluating the classifier with different analysis windows sizes, and demonstrate how smaller window sizes affect the performance of the classifier. The classi- fier is then implemented in a segmentation system. We present the results of an evaluation on three segmentation systems: median filter, k-depth lookahead, and a probabilistic algorithm selecting class association.

Type : Journal Article

Publication : Journal of the Audio Engineering Society, vol. 64, no. 7/8, pp. 484-492, 2016 

 

© 2018 by Jianyu Fan.