Tesla-Rapture: A Lightweight Gesture Recognition System from mmWave Radar Sparse Point Clouds

We present Tesla-Rapture, a gesture recognition system for sparse point clouds generated by mmWave Radars. State of the art models are either too resource consuming or not sufficiently accurate integration into real-life scenarios using wearable constrained equipment such as IoT devices (e.g. Raspberry PI), XR hardware HoloLens), smart-phones. To tackle this issue, we have developed Tesla, Message Passing Neural Network (MPNN) graph convolution approach radar clouds. The model outperforms state on three datasets in terms accuracy while reducing computational complexity and, hence, execution time. In particular, approach, is able to predict almost 8 times faster than most competitor. Our performance evaluation different (environments, angles, distances) shows that Tesla generalizes well and improves up 20% challenging scenarios, through-wall setting sensing at extreme angles. Utilizing develop real-time implementation Radar PI 4 evaluate its time-complexity. also publish source code, trained models, embedded devices.