mmAnomaly: Leveraging Visual Context for Robust Anomaly Detection in the Non-Visual World with mmWave Radar

2026-03-31

AI Summary

This work addresses the challenge of millimeter-wave radar anomaly detection in non-line-of-sight scenarios, where complex non-Gaussian distortions arise due to material properties, clutter, and multipath interference, often causing existing methods—lacking contextual awareness—to misclassify normal signal fluctuations as anomalies. To overcome this limitation, the study introduces, for the first time, RGB-D visual context to guide millimeter-wave anomaly detection. By fusing scene geometry and material semantics, a conditional latent diffusion model is developed to generate visually consistent expected radar spectrograms. A dual-input spectrogram comparison module further enables interpretable anomaly localization. The proposed approach demonstrates significantly enhanced robustness under occlusion, clothing variations, and complex environments, achieving up to 94% F1 score and sub-meter localization accuracy across three real-world applications, while exhibiting strong cross-modal generalization capability.

Abstract

mmWave radar enables human sensing in non-visual scenarios-e.g., through clothing or certain types of walls-where traditional cameras fail due to occlusion or privacy limitations. However, robust anomaly detection with mmWave remains challenging, as signal reflections are influenced by material properties, clutter, and multipath interference, producing complex, non-Gaussian distortions. Existing methods lack contextual awareness and misclassify benign signal variations as anomalies. We present mmAnomaly, a multi-modal anomaly detection framework that combines mmWave radar with RGBD input to incorporate visual context. Our system extracts semantic cues-such as scene geometry and material properties-using a fast ResNet-based classifier, and uses a conditional latent diffusion model to synthesize the expected mmWave spectrum for the given visual context. A dual-input comparison module then identifies spatial deviations between real and generated spectra to localize anomalies. We evaluate mmAnomaly on two multi-modal datasets across three applications: concealed weapon localization, through-wall intruder localization, and through-wall fall localization. The system achieves up to 94% F1 score and sub-meter localization error, demonstrating robust generalization across clothing, occlusions, and cluttered environments. These results establish mmAnomaly as an accurate and interpretable framework for context-aware anomaly detection in mmWave sensing.

Problem

• mmWave radar
• anomaly detection
• non-visual sensing
• context awareness
• signal distortion

Innovation

• mmWave radar
• multimodal anomaly detection
• visual context
• conditional latent diffusion
• through-wall sensing