Blind source separation of hard clipped audio mixtures

Lux, Adahlia Sophie (2023). Blind source separation of hard clipped audio mixtures. University of Birmingham. Ph.D.

Preview

Lux2023PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (2MB) | Preview

Abstract

Blind source separation of hard-clipped audio mixtures is a challenging problem. This thesis presents a novel approach to recovering sources from mixtures affected by hard clipping distortion. This approach relies on making geometric observations about the nature of clipped mixtures. An assumption of disjointness is key. A strict time-disjointness assumption is made in the first two methods. In the final method, we relax this assumption such that only a portion of the time points are time-disjoint. Evaluation experiments demonstrated an increase in source recovery performance for the proposed method when compared with sequential approaches when at least partial time-disjointness was present for speech and synthetic sources.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):