Smart Assistive Technology for People with Visual Field Loss

Younis, Ola
(2019) Smart Assistive Technology for People with Visual Field Loss. PhD thesis, University of Liverpool.

[img] Text
201171738_Dec2019.pdf - Unspecified

Download (65MB) | Preview


Visual field loss results in the lack of ability to clearly see objects in the surrounding environment, which affects the ability to determine potential hazards. In visual field loss, parts of the visual field are impaired to varying degrees, while other parts may remain healthy. This defect can be debilitating, making daily life activities very stressful. Unlike blind people, people with visual field loss retain some functional vision. It would be beneficial to intelligently augment this vision by adding computer-generated information to increase the users' awareness of possible hazards by providing early notifications. This thesis introduces a smart hazard attention system to help visual field impaired people with their navigation using smart glasses and a real-time hazard classification system. This takes the form of a novel, customised, machine learning-based hazard classification system that can be integrated into wearable assistive technology such as smart glasses. The proposed solution provides early notifications based on (1) the visual status of the user and (2) the motion status of the detected object. The presented technology can detect multiple objects at the same time and classify them into different hazard types. The system design in this work consists of four modules: (1) a deep learning-based object detector to recognise static and moving objects in real-time, (2) a Kalman Filter-based multi-object tracker to track the detected objects over time to determine their motion model, (3) a Neural Network-based classifier to determine the level of danger for each hazard using its motion features extracted while the object is in the user's field of vision, and (4) a feedback generation module to translate the hazard level into a smart notification to increase user's cognitive perception using the healthy vision within the visual field. For qualitative system testing, normal and personalised defected vision models were implemented. The personalised defected vision model was created to synthesise the visual function for the people with visual field defects. Actual central and full-field test results were used to create a personalised model that is used in the feedback generation stage of this system, where the visual notifications are displayed in the user's healthy visual area. The proposed solution will enhance the quality of life for people suffering from visual field loss conditions. This non-intrusive, wearable hazard detection technology can provide obstacle avoidance solution, and prevent falls and collisions early with minimal information.

Item Type: Thesis (PhD)
Divisions: Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User: Symplectic Admin
Date Deposited: 16 Jan 2020 10:30
Last Modified: 11 Aug 2021 07:10
DOI: 10.17638/03064504