Automated Radar Data Labeling using MoveNet for Human Gesture Recognition

Integrated radar sensors constitute an emerging technology used from smart home to autonomous driving applications. In combination with machine learning for radar data signal processing, they experience increasing popularity and are heavily investigated. These methods however need huge amounts of labeled training data to perform well in realistic scenarios, which is time consuming and cost intensive. To increase data acquisition speed, this paper presents a system using a fast computer vision model for automated radar data labeling. It is capable of generating 10 labeled radar data frames per second. To demonstrate the abilities it is used for human gesture recognition.