| LDR | | 00000nam u2200205 4500 |
| 001 | | 000000432839 |
| 005 | | 20200224140250 |
| 008 | | 200131s2019 ||||||||||||||||| ||eng d |
| 020 | |
▼a 9781085795357 |
| 035 | |
▼a (MiAaPQ)AAI13809146 |
| 040 | |
▼a MiAaPQ
▼c MiAaPQ
▼d 247004 |
| 082 | 0 |
▼a 004 |
| 100 | 1 |
▼a Wu, Muchen. |
| 245 | 10 |
▼a Context Determination from Cyber-Physical Sensing. |
| 260 | |
▼a [S.l.]:
▼b University of California, Davis.,
▼c 2019. |
| 260 | 1 |
▼a Ann Arbor:
▼b ProQuest Dissertations & Theses,
▼c 2019. |
| 300 | |
▼a 101 p. |
| 500 | |
▼a Source: Dissertations Abstracts International, Volume: 81-04, Section: B. |
| 500 | |
▼a Advisor: Mohapatra, Prasant. |
| 502 | 1 |
▼a Thesis (Ph.D.)--University of California, Davis, 2019. |
| 506 | |
▼a This item must not be sold to any third party vendors. |
| 520 | |
▼a With the wide-spread deployment of sensors in cyber-physical systems, enormous amount of sensing data is generated all the time. Sensors vary from energy-efficient components (e.g. accelerometers, gyroscopes, etc.) in small smart devices to powerful and informative units like cameras and wireless transceivers. Because of the feeds from these sensing capabilities, systems obtain the raw sensing materials from multiple sources to serve as the foundation of making systems intelligent. Traditional devices acquire new observation from sensing sources to allow themselves upgrade to smart devices. Equipped with more sensory information, large systems like smart grids are able to achieve more intelligent functionalities.Yet, the potential of deriving valuable context from disparate sensing inputs remains underdeveloped. It's a critical problem to honor the value of massive sensing data. The research of determining context from cyber-physical sensing faces a few challenges. First, given numerous methods to process a series of sensing data, there are few that facilitate the precise interpretation of context of interest. To extract proper features from different sensors for efficient context determination, the data characterization lays foundation to accurately derive context. Second, it is impossible to form a unified methodology for heterogeneous sensing capabilities, as they contribute to context determination in different formats, levels and aspects. The design of combining multiple modalities of sensing capabilities has a great impact on the determination of context collaboratively. Third, while the versatility of sensing power provides plenty of potential to enhance the intelligence of systems, the overwhelming competence to determine context raises privacy concerns at the same time. It's worth to notice the balance between assisting users' daily life by digging into users' habits and intruding on the user privacy.In this report, my research focuses on the determination of context from cyber-physical sensing capabilities. In the first work, because of the rich information conveyed in an image, the first-person cameras recording plenty of visual data cause risks of privacy leakage when users are in a sensitive environment. The scheme PriFir is proposed to recognize private circumstances by leveraging the less-intrusive sensors (e.g. accelerometers, light sensors, etc.). The determination of sensitive scenarios instructs the on/off of first-person cameras and helps to protect users' private moments from unintentional or malicious exposure. It is shown that PriFir can classify sensitive scenarios with an accuracy of 87% while restraining the false positive rate (percentage of non-sensitive scenarios misclassified as sensitive) to be lower than 5%. The second work targets on the context of building door events to assist monitoring indoor activities and developing safety applications inside the building structures. The context derivation is based on the barometer sensor on an off-the-shelf smartphone only. Despite the presence of user's mobility and the location inside the building, the barometric pressure variation sufficiently determines the building door events and reveals the indoor activities with an accuracy range of 99.34% - 99.81% based on the data collected from 3 different buildings. The third project proposes the framework of fusing multiple sensing modalities (i.e. visual and acoustic sensors) together under the limitation of resource constraint. The fourth work focuses on log data from wireless networks. The processed log-structured data is exploited by recurrent neural networks to uncover the information in another domain, i.e. details of on-campus events. The data acquired from networks benefits the context determination of events in physical world.Research projects reported in this thesis cover variety of cyber-physical sensing sources. The exploration of these sensing sources shares the common life-cycle, i.e. data acquisition, data processing, model formulating, and performance evaluating. These procedures are applied to different scales of applications from personal use, to a single building and to a regional network. The study in this report contributes to: 1) explore the breath of contextual determination from cyber-physical sensing by incorporating variety of sensing modalities |
| 590 | |
▼a School code: 0029. |
| 650 | 4 |
▼a Computer science. |
| 690 | |
▼a 0984 |
| 710 | 20 |
▼a University of California, Davis.
▼b Computer Science. |
| 773 | 0 |
▼t Dissertations Abstracts International
▼g 81-04B. |
| 773 | |
▼t Dissertation Abstract International |
| 790 | |
▼a 0029 |
| 791 | |
▼a Ph.D. |
| 792 | |
▼a 2019 |
| 793 | |
▼a English |
| 856 | 40 |
▼u http://www.riss.kr/pdu/ddodLink.do?id=T15490570
▼n KERIS
▼z 이 자료의 원문은 한국교육학술정보원에서 제공합니다. |
| 980 | |
▼a 202002
▼f 2020 |
| 990 | |
▼a ***1008102 |
| 991 | |
▼a E-BOOK |