CLGNet: A New Network for Human Pose Estimation using Commodity Millimeter Wave Radar

This paper introduces a new network (CLGNet: Combined Local and Global information encoding Network) for human pose estimation based on commodity millimeter wave (mmWave) radar. Based on the benchmark model of ResNet, a global spatial information encoding module is introduced at the early stage of the network. This new module is expected to help learning the relationship between sparsely distributed human pose keypoints with internal relations. Our experimental results show that the addition of this structural module improves the prediction accuracy of human pose keypoints, especially on minor body parts, such as hands and feet.