Terrain awareness mobility model to support outdoor mobility for people with vision impairment

Vision impairment is defined as any degree of impairment to a person’s ability to see that affects his or her daily life. Changes in the ground surface (presence of terrain types) represent a significant challenge for people with vision impairment (PVI). While enormous research work proposed valuable solutions that improved PVI mobility, most of these studies investigated detecting obstacles above the ground and addressing navigation problems, with an insufficient investigation into terrain hazards. In addition to the lack of spatial information provided, which is mostly limited to vibration or audio signals. To address this gap, this study developed a Terrain Awareness Mobility Model (TAM2) to support outdoor mobility for PVI. TAM2 aims to improve terrain awareness by incorporating providing spatial information during mobility. The provided information improves instant mobility performance and facilitates the formation of the cognitive map of the environment, which can support safe and independent outdoor mobility. In this study, the User-Centred Design (UCD) approach was adopted, which involved engaging the target group throughout all research phases. The research design applied the exploratory mixed method, comprising qualitative research, then proposed the solution according to the findings, and finally, evaluated the solution quantitatively. To gather the users’ requirements and expectations, a qualitative study was conducted with two groups of participants. The first group consisted of four experts whose job is to empower the life quality of PVI, while the second group comprised 15 participants with vision impairment. The study utilized three research instruments during this phase, a semi-structured interview, a mobility observation session in a familiar environment, and a mobility observation session in an unfamiliar environment. Thematic analysis was applied, and the findings outlined the primary components of TAM2. TAM2 contains three main components; user model, terrain detection model, and learning model. For the prototyping phase, the study employed the deep learning detection framework YOLOv4-tiny algorithm to implement a real-time terrain detection model. This detection model was integrated with an Android detection app to detect specific types of terrain. Additionally, the app was equipped with a speech message feedback function to convey spatial information, including terrain type, direction, and approximate distance. Furthermore, a real-world quasi-experiment was conducted with 14 participants with vision impairment to evaluate TAM2 effectiveness through the app. The experiment measured the terrain type detection performance and feedback effectiveness, in addition to the usability of the app. Quantitative analysis via the Mann-Whitney U test technique with p-value = 0.05 was applied to assess the mobility performance improvement and cognitive map formation. The analysis revealed a statistically significant improvement in the users’ mobility performance and cognitive map formation for the users who utilized the app. Additionally, the questionnaire descriptive analysis revealed that 71.4% of the participants agreed with the app’s effectiveness and usability. This result indicates that TAM2 is able to support outdoor mobility for PVI. The terrain detection app can be used by the PVI community to improve their terrain awareness. As well the provided measurement methods can assist stakeholders in teaching locomotion for PVI. Moreover, TAM2 can serve as a roadmap for researchers investigating mobility for PVI, outlining the main requirements of the target group and how to achieve these requirements to progress the research. This could lead to the development of ATs that are better suited to PVI requirements (e.g., ATs that complement the white cane) and more usable that improve outdoor mobility for PVI. Accordingly, TAM2 and the terrain detection app represent the key contribution of this study.

Text 119731 (A).pdf Download (13MB) Official URL or Download Paper: http://ethesis.upm.edu.my/id/eprint/18476

Actions (login required)

View Item