One Snapshot is All You Need: A Generalized Method for mmWave Signal Generation

TL;DR

mmGen presents a generalized, software-based framework for mmWave signal generation that synthesizes full-scene reflections—human, environmental, and multipath—using a single environment snapshot. By modeling phase, amplitude, material properties, antenna gains, and HRPP-based multipath, mmGen produces realistic Range-Angle and micro-Doppler signatures that closely match real data (RA MS-SSIM ≈ 0.91, MD MS-SSIM ≈ 0.89). The approach eliminates extensive real-data collection for new motions or environments and demonstrates strong cross-domain transfer in downstream tasks such as activity recognition, achieving 91% accuracy with synthetic training data. Overall, mmGen offers a practical, generalizable tool to boost mmWave sensing research and applications without heavy data collection or task-specific refiners.

Abstract

Wireless sensing systems, particularly those using mmWave technology, offer distinct advantages over traditional vision-based approaches, such as enhanced privacy and effectiveness in poor lighting conditions. These systems, leveraging FMCW signals, have shown success in human-centric applications like localization, gesture recognition, and so on. However, comprehensive mmWave datasets for diverse applications are scarce, often constrained by pre-processed signatures (e.g., point clouds or RA heatmaps) and inconsistent annotation formats. To overcome these limitations, we propose mmGen, a novel and generalized framework tailored for full-scene mmWave signal generation. By constructing physical signal transmission models, mmGen synthesizes human-reflected and environment-reflected mmWave signals from the constructed 3D meshes. Additionally, we incorporate methods to account for material properties, antenna gains, and multipath reflections, enhancing the realism of the synthesized signals. We conduct extensive experiments using a prototype system with commercial mmWave devices and Kinect sensors. The results show that the average similarity of Range-Angle and micro-Doppler signatures between the synthesized and real-captured signals across three different environments exceeds 0.91 and 0.89, respectively, demonstrating the effectiveness and practical applicability of mmGen.

Figures (10)

• Figure 1: The illustration of FMCW chirp and signal preprocessing.
• Figure 2: Overview of mmGen.
• Figure 3: The illustration of how the orientation of a triangle surface affects the reflected signal.
• Figure 4: RA heatmaps before/after performing the static clutter removal.
• Figure 5: Multipath reflection tracing and HRPP table construction.