A Self-adaptable Activity Recognition System for Smart Homes

• Main Authors: Huang, Li-wei, 黃立維
• Other Authors: Alan Liu
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/96832805916805986053

碩士 === 國立中正大學 === 電機工程研究所 === 100 === In order to provide better service for the user in a smart home, how to understand the user's situation and needs becomes an important issue. Knowing the user's activity contributes to the smart home that knows the needs of the user and provides more appropriate service. Because different users may conduct activities in different ways, we cannot completely define all possible activities in advance. This study proposes an activity recognition system by using Bayesian network so that system can make inferences even when context information is uncertain. We have first created separate networks about each user's activity, and then we have combined each network to construct the activity model to complete the activity recognition process. Because each user’s habits and the environment are not the same, we have built a personalized system by using a machine learning technique through collecting training data. In addition, the system has the ability to adjust the model parameters, so that the system can adapt to dynamic environment. Finally, considering that a family often consists of more than one member in general, a smart home needs to support different users who may perform different activities at the same time. Therefore, this thesis presents an algorithm to produce multiple recognition results, so the system can recognize activities performed by multiple users.