Data Augmentation Techniques for Radar-Based Vital Sign Detection

In this study, we propose a deep learning-based data augmentation method to enhance the diversity and quality of radar signal data for vital sign detection. The method combines the addition of random Gaussian noise with the generation of data by deep learning-based generation that closely resembles actual measurements, thereby addressing the challenge of collecting large-scale training datasets. By training deep learning-based classification models on a mixture of real and augmented data, we observe improved classification accuracy compared to models trained without augmentation. These results highlight the potential of the proposed approach in applications such as security, surveillance, and disaster rescue operations.