Acquiring in situ training data for context-aware ubiquitous computing applications
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
IEEE Pervasive Computing
Practical stability of closed-loop descriptor systems
International Journal of Systems Science
IEEE Transactions on Information Technology in Biomedicine
| Hi-index | 0.00 |
This paper deals with a novel system which infers activities of daily living (ADL). Particularly, we present details of component ADL decision. It is a core component in inferring ADL. The target activities are not ambulatory ones (e.g. walk, sit, or lie) but higher level ones such as having a meal, taking a rest, sleeping. The proposed system makes use of a triaxial accelerometer sensor, and environment sensors for indicating a subject's position. The inferring is carried out in real time. And the inferred ADLs are stored at a database immediately in order to be utilized by various applications. We propose a unique ADL inference method which consists of three steps: elementary activity classification, component ADL decision, ADL inference. In particular, the component ADL decision module provides the additional functionality of correcting misclassified elementary activities and refining intermediate component ADLs, which facilitates ADL inferring. These functions improve the accuracy of ADL inference. Preliminary results show that the proposed system effectively infers component ADLs and ultimately ADLs.