Abstract | ||
---|---|---|
The implementation of wearable airbags to prevent fall injuries depends on accurate pre-impact fall detection and a clear distinction between activities of daily living (ADL) and them. We propose a novel pre-impact fall detection algorithm that is robust against ambiguous falling activities. We present a data-driven approach to estimate the fall risk from acceleration and angular velocity features and use thresholding techniques to robustly detect a fall before impact. In the experiment, we collect simulated fall data from subjects wearing an inertial sensor on their waist. As a result, we succeeded in significantly improving the accuracy of fall detection from 50.00 to 96.88%, the recall from 18.75 to 93.75%, and the specificity 81.25 to 100.00% over the baseline method. |
Year | DOI | Venue |
---|---|---|
2021 | 10.1007/978-981-19-0361-8_8 | SENSOR- AND VIDEO-BASED ACTIVITY AND BEHAVIOR COMPUTING, ABC 2021 |
DocType | Volume | ISSN |
Conference | 291 | 2190-3018 |
Citations | PageRank | References |
0 | 0.34 | 0 |
Authors | ||
9 |
Name | Order | Citations | PageRank |
---|---|---|---|
Takuto Yoshida | 1 | 0 | 0.34 |
Kazuma Kano | 2 | 0 | 0.34 |
Keisuke Higashiura | 3 | 0 | 0.34 |
Kohei Yamaguchi | 4 | 0 | 0.34 |
Koki Takigami | 5 | 0 | 0.34 |
Kenta Urano | 6 | 0 | 0.68 |
Shunsuke Aoki | 7 | 0 | 0.68 |
Takuro Yonezawa | 8 | 0 | 0.68 |
Nobuo Kawaguchi | 9 | 0 | 0.68 |