Enhancing access to complex and spatial information for blind users of mobile devices

Thesis (Ph. D.)--University of Rochester. Dept. of Computer Science, 2015.For people with visual impairments, accessing visual information in mobile settings can be difficult and sometimes frustrating. They often need to depend on assistance from family or strangers, and thus limit their exposure to unfamiliar settings so they can avoid inaccessible information. While there are many mobile applications dedicated to providing improved access of visual information for visually impaired users, there is still a long way to go towards providing an ideal user experience. This thesis focuses on enhancing visual information retrieval and interpretation for blind people with two sets of technologies. The first is computer vision, which provides us numerous algorithms to process images and videos in order to efficiently extract visual information of blind users’ surrounding environment, and sometimes fulfill their needs promptly or even in real time. The second is crowdsourcing, also known as human computation. Crowdsourcing has been shown to be an effective means to solve problems that automated approaches cannot yet handle. With carefully designed infrastructure, leveraging large groups of online workers can yield accurate and fast responses to blind users’ various visual information requests. This dissertation surveyed existing algorithms and applications of both computer vision and crowdsourcing and identified challenges visually impaired users face from questions they submitted to our services, then carefully designed frameworks to fit components from both categories together into several assistive mobile applications for blind people