Decoding the Recognition of Occluded Objects in the Human Brain

Mansfield, Courtney Elizabeth (2024) Decoding the Recognition of Occluded Objects in the Human Brain. Doctoral thesis, University of East Anglia.

Preview

PDF Download (6MB) | Preview

Abstract

The dynamics of object recognition are intricate, particularly when under challenging visual conditions, such as occlusion. Current models of vision often fall short in explaining the human visual system's remarkable ability to represent occluded objects. Previous studies have predominantly employed simple shapes as occluders, limiting the understanding of real-world occlusion scenarios. Chapter 2 delves into neural representations by investigating occlusion with realistic stimuli—objects occluding other objects. Using event-related fMRI, participants engaged in a one-back task while viewing objects in isolation, occluded by other objects, or cut out by silhouettes. Decoding analyses in early visual cortex (EVC) revealed a reliance on visible features, while inferotemporal cortex (IT) exhibited robust representations, incorporating both visible and inferred features. Competition effects across multiple objects were evident in EVC but notably weaker in IT, highlighting IT's capacity to disentangle neural responses amidst competing stimuli. Chapter 3 expands the exploration to behavioural aspects, unveiling the impact of occlusion magnitude on recognition difficulty. IT displayed a linear increase of beta weights in processing allocation with recognition difficulty. Behaviourally, unoccluded conditions showed enhanced accuracy and faster response times, with unique recognition patterns emerging when objects served as both occluders and occluded objects. Chapter 4 uses fMRI to examine the theoretical perspective of predictive processing, employing expectation suppression in EVC during occlusion, motivated by high-level occlusion responses found previously. Multivariate pattern analysis indicated an expectation suppression effect aligning with the sharpening account of predictive processing. The concluding chapter synthesises these findings, emphasising the practical and theoretical implications. Notably, the thesis underscores the importance of utilising ecological visual information in visual neuroscience studies and highlights the differing capabilities of EVC and IT. Collectively, our research contributes valuable insights into the neural mechanisms underlying object recognition in challenging visual conditions, paving the way for future research avenues.

Item Type:	Thesis (Doctoral)
Faculty \ School:	Faculty of Social Sciences > School of Psychology
Depositing User:	Chris White
Date Deposited:	06 Aug 2024 07:26
Last Modified:	06 Aug 2024 07:26
URI:	https://ueaeprints.uea.ac.uk/id/eprint/96156
DOI:

Downloads

Actions (login required)

View Item