mmID: High-Resolution mmWave Imaging for Human Identification

TL;DR

This work addresses the challenge of identifying individuals from mmWave images by converting low-resolution spatial spectra into high-resolution body images using a conditional GAN fed with MUSIC-derived spectra. The mmID pipeline jointly reconstructs full-body shapes and then classifies identities with a CNN, achieving 93% accuracy in unseen environments and a 5% mean silhouette difference relative to Kinect ground truth. Key contributions include background removal, MUSIC-based spectrum estimation with joint transmitter smoothing, a 3D-2D multi-stream cGAN discriminator, and a weighted-plus-SSIM perceptual loss design. The results demonstrate the practicality of privacy-preserving, high-resolution mmWave imaging for household- or workplace-scale identification, while highlighting range limitations and hardware leakage as practical constraints. The approach offers a path toward real-world deployment of identification systems based on mmWave sensing under varied environmental conditions.

Abstract

Achieving accurate human identification through RF imaging has been a persistent challenge, primarily attributed to the limited aperture size and its consequent impact on imaging resolution. The existing imaging solution enables tasks such as pose estimation, activity recognition, and human tracking based on deep neural networks by estimating skeleton joints. In contrast to estimating joints, this paper proposes to improve imaging resolution by estimating the human figure as a whole using conditional generative adversarial networks (cGAN). In order to reduce training complexity, we use an estimated spatial spectrum using the MUltiple SIgnal Classification (MUSIC) algorithm as input to the cGAN. Our system generates environmentally independent, high-resolution images that can extract unique physical features useful for human identification. We use a simple convolution layers-based classification network to obtain the final identification result. From the experimental results, we show that resolution of the image produced by our trained generator is high enough to enable human identification. Our finding indicates high-resolution accuracy with 5% mean silhouette difference to the Kinect device. Extensive experiments in different environments on multiple testers demonstrate that our system can achieve 93% overall test accuracy in unseen environments for static human target identification.

Figures (6)

• Figure 1: mmID System Design
• Figure 2: Human Body Imaging Reconstruction Network
• Figure 3: Data collection devices and setup. (a)60GHz Qualcomm device with co-located Tx and Rx, (b) $6\times 6$ antenna layout, $\theta$ and $\phi$ denotes the elevation and azimuth angle respectively, (c) Data collection setup with mmWave device and Kinect.
• Figure 4: Human body imaging performance. (a) mmWave spectrum estimated using MUSIC for one Tx, (b) Generated image, and (c) Ground truth image obtained using Kinect.
• Figure 5: Classification accuracy confusion matrix for unseen environment