Locating senior walking frame users in crowded indoor environments
- Publication Type:
- Thesis
- Issue Date:
- 2017
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
This thesis presents a low cost indoor localization system, primarily intended for use by professional elder care supervisors for tracking elderly people in their excursions to a crowded shopping centre. The main requirement is that the system provides an approximate locations of multiple elderly people during an excursion to a crowded shopping centre. The residents are to use walking frames for locomotion, thus their motion is relatively slow and predictable. The resolution of localization is considered adequate if the care supervisor is able to locate a given person through visual contact relative to the estimated location. This thesis presents two novel localization methods that make use of these simplifying constraints and provides an industry strength implementation of one of these strategies.
The first method described is an image based place recognition technique that employs the Bag of Words model for generating image descriptors and a three layer feedforward neural network for producing location estimates. Shop fronts and their corresponding neighbourhood areas are used as classes for training the neural network. The performance of this approach that was evaluated in a real shopping centre environment is presented. Although the system developed performs well, it was found to require the user cooperation in crowded areas and was deemed to have potential privacy concerns.
An alternative solution, a Wi-Fi based indoor localization method is also presented. It estimates the current location of a subject using the Wi-Fi signal strengths received by a sensor module mounted on a walking frame. The environment is modelled as a collection of cells with sizes sufficiently small for locating a person through eye contact. A motion model, based on the knowledge of the floor plan of the environment is described. A probabilistic framework using the Bayes rule in combination with a Kernel Density method for estimating the probability density functions of received signal strengths at the cells is developed.
The Wi-Fi based indoor localization method was implemented on a unit that measures strengths of Wi-Fi signals received from the access points present in an environment, computes the location and transmits it using the telephone network to a tablet held by a carer. An application on the tablet for visualizing the location of multiple walkers was also developed. The performance of this system was evaluated by conducting multiple trials, including a shopping centre excursion organized by a professional elder care provider named IRT.
From the localization accuracies obtained through the test trials, it could be concluded that the presented Wi-Fi based localization method is adequate to fulfil the requirement of IRT, which is locating elderly people in a crowded indoor environment. It is also found that the floor plan based motion model enables the localization algorithm to produce reliable location estimates, given the relatively slow motion of elderly people.
Please use this identifier to cite or link to this item: