In a practical setting of radar-based fall detection system, radar data arrives as a continuous, stream rather than packaged events. Before the system can recognise a fall, it must first constantly scan this incoming stream to isolate the exact moments of genuine movement from empty background recordings. The conventional approach of using a fixed energy threshold to make this approach vulnerable to failure because it cannot adapt to its unexpected background noise. This situation can easily trigger false alarms, and on the other hand, weak signals from a distant fall might go completely undetected. Furthermore, a threshold perfectly tuned for one room often fails in another, making constant manual recalibration impractical for real-world deployment to address this, we propose a noise-adaptive diversity-combining approach. Based on the principle of maximum-ratio combining, this method weights horizontal and vertical radar energies according to their respective noise levels. We evaluated our approach on a dual-radar dataset against independent time-domain ground truth, testing across simulated signal-to-noise ratios (SNRs) from +20 to −10 dB. As expected, uncalibrated fixed-thresholds performed poorly (mean F1-score: 0.095). However, our proposed method (F1-score: 0.75–0.86) also significantly outperformed calibrated single- and dual-radar constant false-alarm-rate (CFAR) baselines (F1-score: 0.74–0.77) at every SNR level (p < 0.001). These results demonstrate that threshold calibration alone is insufficient for robust frame detection, and that our noise-adaptive approach provides consistent improvements regardless of environmental noise
| Item Type: | Article |
|---|---|
| Creators: | Creators Email / ID Num. Syahirah, Nur Iman UNSPECIFIED Abdul Rashid, Nur Emileen UNSPECIFIED Awang Kechik, Mohd Mustafa UNSPECIFIED Mohd Shariff, Khairul Khaizi UNSPECIFIED |
| Subjects: | T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Telecommunication T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Radar |
| Divisions: | Universiti Teknologi MARA, Shah Alam > Faculty of Electrical Engineering |
| Journal or Publication Title: | Journal of Electrical and Electronic Systems Research (JEESR) |
| UiTM Journal Collections: | UiTM Journals > Journal of Electrical and Electronic Systems Research (JEESR) |
| ISSN: | e-ISSN : 3030-640X |
| Volume: | 29 |
| Number: | 1 |
| Page Range: | pp. 250-257 |
| Keywords: | Fall detection, Dual-radar system, CFAR detection, Diversity combining, Active frame detection |
| Date: | 31 October 2026 |
| URI: | https://ir.uitm.edu.my/id/eprint/149179 |
149179.pdf
