RadHARSimulator V1: Model-Based FMCW Radar Human Activity Recognition Simulator
Fuente:
arXiv
Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912580340023296 |
|---|---|
| author | Gao, Weicheng |
| author_facet | Gao, Weicheng |
| contents | Radar-based human activity recognition (HAR) is a pivotal research area for applications requiring non-invasive monitoring. However, the acquisition of diverse and high-fidelity radar datasets for robust algorithm development remains a significant challenge. To overcome this bottleneck, a model-based frequency-modulated continuous wave (FMCW) radar HAR simulator is developed. The simulator integrates an anthropometrically scaled $13$-scatterer kinematic model to simulate $12$ distinct activities. The FMCW radar echo model is employed, which incorporates dynamic radar cross-section (RCS), free-space or through-the-wall propagation, and a calibrated noise floor to ensure signal fidelity. The simulated raw data is then processed through a complete pipeline, including moving target indication (MTI), bulk Doppler compensation, and Savitzky-Golay denoising, culminating in the generation of high-resolution range-time map (RTM) and Doppler-time maps (DTMs) via both short-time Fourier transform (STFT) and Fourier synchrosqueezed transform (FSST). Finally, a novel neural network method is proposed to validate the effectiveness of the radar HAR. Numerical experiments demonstrate that the simulator successfully generates high-fidelity and distinct micro-Doppler signature, which provides a valuable tool for radar HAR algorithm design and validation. The installer of this simulator is released at: https://github.com/JoeyBGOfficial/RadHARSimulatorV1-Model-Based-FMCW-Radar-Human-Activity-Recognition-Simulator. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_06751 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | RadHARSimulator V1: Model-Based FMCW Radar Human Activity Recognition Simulator Gao, Weicheng Signal Processing 68T45 I.5.4 Radar-based human activity recognition (HAR) is a pivotal research area for applications requiring non-invasive monitoring. However, the acquisition of diverse and high-fidelity radar datasets for robust algorithm development remains a significant challenge. To overcome this bottleneck, a model-based frequency-modulated continuous wave (FMCW) radar HAR simulator is developed. The simulator integrates an anthropometrically scaled $13$-scatterer kinematic model to simulate $12$ distinct activities. The FMCW radar echo model is employed, which incorporates dynamic radar cross-section (RCS), free-space or through-the-wall propagation, and a calibrated noise floor to ensure signal fidelity. The simulated raw data is then processed through a complete pipeline, including moving target indication (MTI), bulk Doppler compensation, and Savitzky-Golay denoising, culminating in the generation of high-resolution range-time map (RTM) and Doppler-time maps (DTMs) via both short-time Fourier transform (STFT) and Fourier synchrosqueezed transform (FSST). Finally, a novel neural network method is proposed to validate the effectiveness of the radar HAR. Numerical experiments demonstrate that the simulator successfully generates high-fidelity and distinct micro-Doppler signature, which provides a valuable tool for radar HAR algorithm design and validation. The installer of this simulator is released at: https://github.com/JoeyBGOfficial/RadHARSimulatorV1-Model-Based-FMCW-Radar-Human-Activity-Recognition-Simulator. |
| title | RadHARSimulator V1: Model-Based FMCW Radar Human Activity Recognition Simulator |
| topic | Signal Processing 68T45 I.5.4 |
| url | https://arxiv.org/abs/2509.06751 |