Thesis
Comparing sensor modalities for activity recognition
Washington State University
Master of Science (MS), Washington State University
2011
Handle:
https://hdl.handle.net/2376/105702
Abstract
Activity recognition has a promising role in various applications such as health care, psychology, and security. Choosing an appropriate sensing modality to gather data is one of the most important factors in effective activity recognition. The sensory modality has an impact both on the final results and also on the degree to which users will accept the technology. In this work we evaluate different sensing technologies (Environmental, Object and Wearable) for the purpose of activity recognition in smart environments based on the type of activities being recognized. In this thesis, we introduce different sensing technologies and discuss both the positive points and the limitations of each. We also conduct experiments in a real home setting with participants performing common activities of daily living. Alternative data features and activity recognition algorithms are tested with the goal of determining an optimal sensor class for a type of activity. We discuss results based on different sensor combinations and provide suggestions about which sensor technology is most suitable for recognizing a particular class of activity. In addition, this study introduces the notion of a suffix tree to adapt pattern discovery techniques to the problem of activity recognition with wearable sensors. This model is evaluated using data gathered from wearable accelerometers. Finally, we present a formal analysis of activity complexity. By defining measurements in terms of three dimensions, sensing, computational and performance, this analysis characterizes activities in terms of a complexity measure. Moreover, we introduce grammars as a formal representation of activities and propose such grammars as an approach for measuring the complexity of an activity.
Metrics
20 File views/ downloads
52 Record Views
Details
- Title
- Comparing sensor modalities for activity recognition
- Creators
- Yasamin Sahaf
- Contributors
- Diane J. Cook (Degree Supervisor)
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University; Pullman, Wash. :
- Identifiers
- 99900525009601842
- Language
- English
- Resource Type
- Thesis