Abstract
Human motion sensing plays a crucial role in smart systems fordecision-making, user interaction, and personalized services. Extensiveresearch that has been conducted is predominantly based on cameras, whoseintrusive nature limits their use in smart home applications. To address this,mmWave radars have gained popularity due to their privacy-friendly features. Inthis work, we propose milliFlow, a novel deep learning approach to estimatescene flow as complementary motion information for mmWave point cloud, servingas an intermediate level of features and directly benefiting downstream humanmotion sensing tasks. Experimental results demonstrate the superior performanceof our method when compared with the competing approaches. Furthermore, byincorporating scene flow information, we achieve remarkable improvements inhuman activity recognition and human parsing and support human body parttracking.