Please use this identifier to cite or link to this item:
Type: Thesis
Title: Ambulatory Monitoring Using Passive RFID Technology
Author: Wickramasinghe, Wickramasinghe Mudiyanselage Asanga Sampath Bandara
Issue Date: 2017
School/Discipline: School of Computer Science
Abstract: Human activity recognition using wearable sensors is a growing field of study in pervasive computing that forms the basis for ubiquitous applications in areas like health care, manufacturing, human computer interaction and sports. A new generation of passive (batteryless) sensors such as sensor enabled RFID (Radio Frequency Identification) tags are creating new prospects for wearable sensor based applications. As passive sensors are lightweight and small, they can be used for unobtrusive monitoring. Furthermore, these sensors are maintenance free as they require no battery. However, recognising activities from passive sensor enabled RFID tags is challenging due to the sparse and noisy nature of the data streams from these sensors because they need to harvest adequate energy for successful operation. Therefore, within this thesis, we propose methods to recognise activities in real time using passive RFID technology by alleviating the adverse effects of sparsity and noise. We mainly consider ambulatory monitoring to facilitate mitigating falls in hospitals and older care settings as our application context. Specifically, three aspects are considered: i) data acquisition from sensor enabled RFID tags; ii) monitoring ambulatory movements using passive sensor enabled RFID tags to recognise activities leading to falls; and iii) detecting falls using a dense deployment of passive RFID tags. A generic middleware architecture and a generic tag ID format to embed sensor data and uniquely identify tag capabilities are proposed to acquire sensor data from passive sensor enabled RFID tags. The characteristics of this middleware are established using experiments with RFID readers and an example application scenario. In the context of ambulatory monitoring using passive sensor enabled RFID tags, first, an algorithm to facilitate the online interpolation of sparse accelerometer data from passive sensor enabled RFID tags is proposed followed by an investigation of features for activity recognition. Secondly, two data stream segmentation methods are proposed that can segment the data stream on possible activity boundaries to mitigate the adverse effects posed by data stream sparsity on segmentation. Thirdly, an algorithm to model the sequential nature considering previous sensor observations for a given time and their class labels to classify a sparse data stream in real time is proposed. Finally, a classification algorithm based on structured prediction is proposed to both segment and classify the sensor data stream simultaneously. The proposed methods are evaluated using four datasets that have been collected from a passive sensor enabled RFID tag with an accelerometer and successful monitoring of ambulatory movements is demonstrated to be possible by employing innovative data stream processing methods, based on machine learning. In order to detect falls, particularly long lie situation, using a dense deployment of passive RFID tags embedded in a carpet, an efficient and scalable machine learning based algorithm is proposed. This algorithm relies only on binary tag observation information. First, it identifies possible fall locations using heuristics and then the falls are identified using machine learning from features extracted considering possible fall locations alone. From an evaluation, it is demonstrated that the proposed algorithm could successfully identify falls in real time.
Advisor: Neumann, Frank
Dissertation Note: Thesis (Ph.D.) -- University of Adelaide, School of Computer Science, 2017
Keywords: Activity recognition
sparse data streams
machine learning
passive sensor enabled RFID
older people
Provenance: This electronic version is made publicly available by the University of Adelaide in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the owner of any included third party copyright material you wish to be removed from this electronic version, please complete the take down form located at:
Appears in Collections:Research Theses

Files in This Item:
File Description SizeFormat 
Wickramasinghe2017_PhD.pdf8.77 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.