|
|
Navigation and Information System for Visually Impaired
Hrbáček, Jan ; Volf, Jaromír (oponent) ; Grepl, Robert (oponent) ; Singule, Vladislav (vedoucí práce)
Visual impairment is one of the most widespread physical handicaps -- as much as 3 % of the world's population suffers from visual impairment or complete blindness. Vision loss substantially worsens one's ability to orient in the environment -- without the knowledge of spatial arrangement, normally acquired predominantly by one's vision, the impaired simply does not know which way to go. The most usual solution is an accompanying person; this service is very demanding though and the sightless have to fully rely on the accompaniment. This thesis explores ways to support visually impaired users' orientation in space by employment of existing sensory technology and application of appropriate processing methods. The subject is researched through an analogy to mobile robotics, by partitioning the subject to the separate problems of localization and path planning. While the methods of path planning are generally existing, pedestrian localization often suffers from major inaccuracies and complicates usage of standard navigation devices by the visually impaired users. An improvement of pose estimate quality can be accomplished by a multitude of approaches studied by the analytical section. In the first stage, the thesis proposes data fusion between an ordinary GPS receiver and a pedestrian dead reckoning unit, leading to preserving local trajectory shape feature faithfully. To mitigate remaining offset errors, design of a globally referenced mechanism of natural landmark detection and matching is provided. Building on the existing graph search formalism, the path planning part of the thesis examines optimality criteria suitable for navigating the visually impaired user through urban terrain. A human-oriented fuzzy logic driven high-level instructions generator is then devised together with a real-time haptic feedback delivering heading directions. The performance of the proposed techniques was evaluated in real-world scenarios, aiming to capture the particularities of the target urban environment. The outcomes have shown considerable improvements in both maximum and mean positioning errors.
|
|
|
Navigation and Information System for Visually Impaired
Hrbáček, Jan ; Volf, Jaromír (oponent) ; Grepl, Robert (oponent) ; Singule, Vladislav (vedoucí práce)
Visual impairment is one of the most widespread physical handicaps -- as much as 3 % of the world's population suffers from visual impairment or complete blindness. Vision loss substantially worsens one's ability to orient in the environment -- without the knowledge of spatial arrangement, normally acquired predominantly by one's vision, the impaired simply does not know which way to go. The most usual solution is an accompanying person; this service is very demanding though and the sightless have to fully rely on the accompaniment. This thesis explores ways to support visually impaired users' orientation in space by employment of existing sensory technology and application of appropriate processing methods. The subject is researched through an analogy to mobile robotics, by partitioning the subject to the separate problems of localization and path planning. While the methods of path planning are generally existing, pedestrian localization often suffers from major inaccuracies and complicates usage of standard navigation devices by the visually impaired users. An improvement of pose estimate quality can be accomplished by a multitude of approaches studied by the analytical section. In the first stage, the thesis proposes data fusion between an ordinary GPS receiver and a pedestrian dead reckoning unit, leading to preserving local trajectory shape feature faithfully. To mitigate remaining offset errors, design of a globally referenced mechanism of natural landmark detection and matching is provided. Building on the existing graph search formalism, the path planning part of the thesis examines optimality criteria suitable for navigating the visually impaired user through urban terrain. A human-oriented fuzzy logic driven high-level instructions generator is then devised together with a real-time haptic feedback delivering heading directions. The performance of the proposed techniques was evaluated in real-world scenarios, aiming to capture the particularities of the target urban environment. The outcomes have shown considerable improvements in both maximum and mean positioning errors.
|
host ::
RSS.