Pattern learning in smart homes and offices using motion sensor and mind wave data : unsupervised approaches
Abstract/Contents
- Abstract
- The general purpose of smart homes and smart offices is to provide people with personalized experiences according to their behaviors and intentions. The foundation of a smart home or office system involves the sensing, learning and understanding of human behaviors. Nowadays, a wide variety of sensors are available on the market that capture information regarding people's physical and mental activities in real-time. The key challenge is to extract useful information from this rich sensor data, in order to learn and recognize people's behavioral patterns. This dissertation focuses on learning and recognizing people's behavioral patterns. It explores people's physical and mental behaviors separately, and presents multiple unsupervised pattern learning approaches for extracting and recognizing these behaviors. The first half of the dissertation concentrates on learning people's movement patterns using motion sensors. Specifically, I first present an activation model that can be used to interpret motion sensor data as a sequence of people's movement events. Based on the activation model, I propose a trajectory-based pattern learning algorithm and develop a decomposition-based pattern learning method, both of which can be used to learn and extract a person's movement patterns from the unlabeled motion sensor data without human input. To validate both methods, the Aruba data set is used as a training set and its activity annotations are regarded as ground truth. The result shows high consistency between the output from both approaches and the activity annotations of the data set. The second half of the dissertation focuses on extracting people's mental workload patterns using MindWave sensors. In particular, I propose the Mental Motion State Model (MMSM) to extract the mental workload patterns. By applying the MMSM model on a MindWave sensor data set collected from 13 subjects during online project meetings in a global teamwork course, three mental motions states are discovered: mental deep recovery, mental light recovery, and mental running. Based on the MMSM model, the mental recovery curve model is introduced which shows objective insights about how people recover and increase their mental energy at different mental workload levels. Using the pattern learning methods proposed in this dissertation together, people's movement patterns and mental workload patterns can be extracted and learned from the motion sensor data and MindWave data without human supervision. Therefore, the work presented in this dissertation can serve as the foundation of an adaptive system, which is able to provide people with automatic assistances and feedback according to the observed behavior patterns.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2016 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Zhang, Tongda |
|---|---|
| Associated with | Stanford University, Department of Electrical Engineering. |
| Primary advisor | Fischer, Martin, 1960 July 11- |
| Thesis advisor | Fischer, Martin, 1960 July 11- |
| Thesis advisor | Fruchter, Renate |
| Thesis advisor | Poon, Ada Shuk Yan |
| Thesis advisor | Tobagi, Fouad A, 1947- |
| Advisor | Fruchter, Renate |
| Advisor | Poon, Ada Shuk Yan |
| Advisor | Tobagi, Fouad A, 1947- |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Tongda Zhang. |
|---|---|
| Note | Submitted to the Department of Electrical Engineering. |
| Thesis | Thesis (Ph.D.)--Stanford University, 2016. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2016 by Tongda Zhang
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...