Text2Doppler: Generating Radar Micro–Doppler Signatures for Human Activity Recognition via Textual Descriptions

Radar-based human activity recognition (HAR) is popular because of its privacy and contactless sensing capabilities. However, a major challenge in this area is the lack of large and diverse datasets. In response, we present a novel framework that uses generative models to transform textual descriptions into motion data, thereby simulating radar signals. This approach significantly enriches the realism and diversity of the dataset, especially for infrequent but critical activities, such as falls and abnormal walking. Textual descriptions capture the semantic complexity of human actions, thereby improving intraclass diversity. Our framework scales the data generation process by using a lightweight physics-based simulator and improves diversity by controlling gait variation, multiviewpoint adaptation, and background noise modeling. The experiments show that data diversity is a critical factor for fair model comparisons, and that the simulated data can effectively improve performance through sim-to-real transfer learning.