Skip to main content
School of Electronic Engineering and Computer Science

Louise Thorpe



Project title: Using Signal-informed Source Separation (SISS) principles to improve instrument separation from legacy recordings 

C4DM theme affiliation: Music Informatics 

Abstract: This project aims to use one audio signal to inform the separation of an audio mixture. The informing signal will be used to create a caricature of a similar signal within the mixture and, by using features of the informing signal, we may extract the similar signal. This research can be positioned alongside existing research that aims to achieve accurate source separation. The main objective of the research will be to explore and evaluate how well an informing signal can be used to obtain a corresponding similar signal and how similar to the target source this must be. By using deep learning, the problems of separating complex musical signals can be addressed. 


Back to top