đī¸ MM Wave Radar Human Identification
This research outlines the critical architecture and methodology needed to achieve individual human identification (identifying which person is present) using mmWave radar, based on the principles demonstrated by the TI Sensor-to-Cloud project.
đī¸ Open-Set Gait Recognition from Sparse mmWave Radar Point Clouds
This work tackles the Open-Set Gait Recognition problem using sparse mmWave radar point clouds, the system must classify known individuals and detect unknown ones. The data is noisy and sparse, unlike dense micro-Doppler spectrograms used in most past work. It solves the realistic open-set human identification problem through a multi-branch adversarial autoencoder (PCAA) that fuses classification, reconstruction, and probabilistic detection. It demonstrates the ability to handle unseen subjects while running on edge-friendly radar data.