Files and links (1)
CC BY-NC V4.0, Open Access
Thesis
Energy-Efficient Wearable Activity Recognition through Activity-Aware Sensor Data Compression and Exploring the Usage of Ultrawide-band Radars for HAR
Washington State University
Master of Science (MS), Washington State University
2023
DOI:
https://doi.org/10.7273/000005116
Abstract
Human activity recognition (HAR) using low-power wearable devices provides an attractive solution for several applications, including remote health monitoring, rehabilitation, and activity promotion. Recent approaches for HAR collect data from multiple sensors mounted on the body to improve the accuracy or the number of activities being recognized. Multiple sensors necessitate communication between the sensors and the processing nodes, leading to high energy consumption. However, transmitting raw sensor data over wireless communication technologies such as WiFi or BLE may not be feasible on energy-constrained wearable devices. This work proposes a novel activity-aware sensor data compression and approximation approach for neural-network-based HAR classifiers. Our approach is based on two key insights: 1) most daily human activities do not have high data variations, thus allowing high compression ratios, and 2) neural network classifiers inherently include some redundancy and can be trained to handle approximations in sensor data. Our experiments
on three datasets show that the proposed approach achieves an average compression ratio of 14 times with as much as 85–95% communication energy savings over a day for a typical pattern of human activities compared to no compression scenario. Moreover, we perform the exploratory study on using Ultra-Wideband (UWB) for HAR problems, show the technical possibilities and limitations, discuss previous works targeting healthcare-related applications, and perform preliminary experiments to show the usage of UWB for HAR.
Metrics
70 File views/ downloads
17 Record Views
Details
- Title
- Energy-Efficient Wearable Activity Recognition through Activity-Aware Sensor Data Compression and Exploring the Usage of Ultrawide-band Radars for HAR
- Creators
- Andrii Zhuravchak
- Contributors
- Ganapati Bhat (Advisor)Hyung Gyu Lee (Committee Member)Monowar Hasan (Committee Member)Yan Yan (Committee Member)
- 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
- Number of pages
- 76
- Identifiers
- 99901019936001842
- Language
- English
- Resource Type
- Thesis