Reed-Jones, J, Marsland, J, Ellis, D, Fergus, P and Jones, K (2023) Early vs Late Fusion in Binaural Sound Source Localisation using CNN. In: International Conference on Intelligent Systems and New Applications (ICISNA'23) Proceedings Book . (International Conference on Intelligent Systems and New Applications (ICISNA'23), Liverpool, UK).

Preview

Text ICISNA_Jago_KOJcomments.pdf - Accepted Version Download (249kB) | Preview

Abstract

In Binaural Sound Source Localisation there are two representations of the signals which contain useful cues for localisation: the time/phase frequency spectrum and the magnitude frequency spectrum. This typically leads to two branch CNN architectures being employed achieve localisation.
This paper compares the difference in performance between models which employ early and later fusion of these two branches, finding only negligible differences and thus concluding that this is an unimportant consideration in the design of such systems.

Item Type:	Conference or Workshop Item (Paper)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > T Technology (General) T Technology > TA Engineering (General). Civil engineering (General)
Divisions:	Computer Science & Mathematics Engineering
Publisher:	ICISNA
SWORD Depositor:	A Symplectic
Date Deposited:	19 Jun 2023 10:42
Last Modified:	19 Jun 2023 10:42
URI:	https://researchonline.ljmu.ac.uk/id/eprint/19435

View Item