ICA 2007

ICA 2007
7th International Conference on
Independent Component Analysis
and Signal Separation

London, UK        9 - 12 September 2007

Banner showing images of London
- Home
- Committee
- Call for Papers
- Submission
- Info for Presenters
- Dates
- Programme
- Tutorials
- Keynotes
- Papers
- Registration
- Accommodation
- Venue
- Maps
- Arrival
- Travel Tips
- Links
- Contact

Paper No: 126

Modeling Perceptual Similarity of Audio Signals for Blind Source Separation

Author(s): Brendan Fox, Andrew Sabin, Bryan Pardo, Alec Zopf

Abstract

Existing perceptual models of audio quality, such as PEAQ, were designed to measure audio codec performance and are not ideally suited to evaluation of audio source separation algorithms. We propose to create a model that focuses on the perception of the errors associated with source separation algorithms, when applied to musical signals. To this end, we collected subjective human assessments of the types of distortions encountered in BASS applications from 31 people. We then mapped these human assessments onto 14 machine-measurable parameters, both with neural networks and linear regression. Results show a strong correlation between a linear combination of a subset of these parameters and mean human assessments. This correlation is much stronger than the correlation between human assessments and several measures currently in use.

Last Updated: 14-Aug-2007   Please read our disclaimer