Behavior modeling in smart environments using camera networks
Abstract/Contents
- Abstract
- The goal of smart environments is to support people in their daily life by being sensitive and responsive to their presence and activities. The key challenges in achieving such a vision are the ability to robustly extract useful information from the sensors, and to utilize the extracted data to generate and reinforce a model representing the user's pattern of behavior so adaptive, personalized or even predictive responses can be provided. In the first part of the dissertation, we discuss how information about people in an environment can be obtained using a network of cameras. We propose efficient visual processing algorithms to track multiple people in real-time, to determine their pose and to estimate their head orientation. A discriminative classifier is trained to detect interactions between occupants of a space based on their locations and head orientations. We also describe methods to fuse information at different levels in a multi-camera scenario. In the second part of the dissertation, we discuss how behavior models can be constructed from observations obtained by cameras or other sensors. We propose a probabilistic model that takes into account the inherent hierarchical structure of human activities, as well as contextual information related to the user or the environment. Our model also allows for variations in the duration of different activities, and considers the uncertainty of the measurements reported by the sensors. To evaluate our visual processing algorithms and illustrate the benefits of our proposed behavior modeling approach, we study three use cases. In the first use case, we study the daily routine of an office worker, and derive the user's pattern of work. In the second use case, our proposed method is applied to a smart home to adaptively optimize energy consumption and user comfort by learning user preferences and habits. Finally, we consider a scenario where the built-in webcam on a personal computer is used to analyze the user's posture and gaze while working on the computer. Personalized ergonomic recommendations tailored to the user's behavior pattern are then provided. Experimental results are presented in each of the use cases to demonstrate the performance of our proposed methods.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2012 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Chen, Zhiwei |
|---|---|
| Associated with | Stanford University, Department of Electrical Engineering |
| Primary advisor | Meng, Teresa H |
| Thesis advisor | Meng, Teresa H |
| Thesis advisor | Aghajan, Hamid K |
| Thesis advisor | El Gamal, Abbas A |
| Advisor | Aghajan, Hamid K |
| Advisor | El Gamal, Abbas A |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Chih-Wei Chen. |
|---|---|
| Note | Submitted to the Department of Electrical Engineering. |
| Thesis | Thesis (Ph.D.)--Stanford University, 2012. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2012 by Chih-Wei Chen
- 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...