摘要
With the rapid development of IoT technologies and widespread deployment of intelligent sensing devices, dynamic sensor selection has become crucial for balancing recognition accuracy and energy efficiency in applications like health monitoring. Existing runtime assessment methods suffer from limited predictive utility for future sensor quality and fail to capture dynamic spatio-temporal dependencies among interconnected devices. To this end, we propose a utility-aware sensor selection framework that integrates adaptive spatio-temporal forecasting with attention-based quality assessment. Our approach employs a Dynamic Graph Learning-based Spatio-Temporal Graph Convolutional Network (DGL-STGCN) to model evolving device correlations and predict sensor states, while an attention mechanism assesses future sensor utility through learned importance scores. This enables proactive sensor selection optimized for both recognition accuracy and energy consumption, explicitly accounting for switching overheads. We develop and evaluate a prototype system designed for wearable devices to monitor locomotion activities. Results show that our system achieves a 14% accuracy improvement over single-sensor baselines, a 25% reduction in energy consumption compared to full-sensor deployment, and a 2% higher accuracy than state-of-the-art methods at equivalent energy levels.
| 源语言 | 英语 |
|---|---|
| 主期刊名 | Harmonious Human-Machine Environment - 21st National Joint Conference, HHME 2025, Proceedings |
| 编辑 | Shihui Guo, Ye Tao, Zhu Wang |
| 出版商 | Springer Science and Business Media Deutschland GmbH |
| 页 | 84-103 |
| 页数 | 20 |
| ISBN(印刷版) | 9789819239931 |
| DOI | |
| 出版状态 | 已出版 - 2027 |
| 活动 | 21st National Joint Conference on Harmonious Human-Machine Environment, HHME 2025 - Dalian, 中国 期限: 21 8月 2025 → 25 8月 2025 |
丛书
| 姓名 | Communications in Computer and Information Science |
|---|---|
| 卷 | 2879 CCIS |
| ISSN(印刷版) | 1865-0929 |
| ISSN(电子版) | 1865-0937 |
会议
| 会议 | 21st National Joint Conference on Harmonious Human-Machine Environment, HHME 2025 |
|---|---|
| 国家/地区 | 中国 |
| 市 | Dalian |
| 时期 | 21/08/25 → 25/08/25 |
联合国可持续发展目标
此成果有助于实现下列可持续发展目标:
-
可持续发展目标 7 经济适用的清洁能源
学术指纹
探究 'Utility-Aware Runtime Assessment of Sensor Selection Based on Spatio-Temporal Dynamic Graph Learning' 的科研主题。它们共同构成独一无二的学术指纹。引用此
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver