| DOI | Resolve DOI: https://doi.org/10.1109/LSENS.2023.3284652 |
|---|
| Author | Search for: Dey, AnkitaORCID identifier: https://orcid.org/0000-0002-5201-1081; Search for: Rajan, SreeramanORCID identifier: https://orcid.org/0000-0003-0153-6723; Search for: Xiao, Gaozhi1ORCID identifier: https://orcid.org/0000-0001-7717-1818; Search for: Lu, Jianping1ORCID identifier: https://orcid.org/0000-0003-3152-7510 |
|---|
| Affiliation | - National Research Council Canada. Quantum and Nanotechnologies
|
|---|
| Format | Text, Article |
|---|
| Subject | Sensor signal processing; fall; one-class classification (OCC); pretrained models; radar; spectrograms; feature extraction; training; event detection; monitoring; sensors |
|---|
| Abstract | Existing deep learning and machine learning techniques for radar-based fall event detection require a large number of fall activity instances at the training stage to produce reliable results. However, to replicate real-life scenarios, falls should be treated as rare events. This work pioneers the application of one-class classification (OCC) technique for radar-based fall event detection. A comparative analysis of three different types of OCC techniques using abstract features extracted from six different pretrained models is shown. An improved novel feature fused OCC model is proposed that achieves an area under the curve value of 0.912 for fall event detection. All evaluations are done on a publicly available dataset of human radar signatures. |
|---|
| Publication date | 2023-06-09 |
|---|
| Publisher | IEEE |
|---|
| In | |
|---|
| Language | English |
|---|
| Peer reviewed | Yes |
|---|
| Export citation | Export as RIS |
|---|
| Report a correction | Report a correction (opens in a new tab) |
|---|
| Record identifier | f3c991fa-d66d-4900-b41b-1329e17a61f1 |
|---|
| Record created | 2024-10-21 |
|---|
| Record modified | 2024-10-21 |
|---|
