Deep learning-based elderly gender classification using Doppler radar

Society today is facing a rapidly aging population. While various monitoring systems have been proposed for protecting elderly persons in their daily lives, concerns relating to privacy limit the effectiveness of these systems. In response to this issue, we investigate the use of Doppler radar images for monitoring the elderly, as these images are known to protect privacy very well. As the first step, we investigate the use of Doppler radar images for the gender classification of the elderly. We used sit-to-stand Doppler radar images of elderly persons, obtained eleven groups of images through image processing, and applied five state-of-the-art deep learning models to classify the gender. The classification results revealed a classification accuracy rate as high as 90%, which indicates that sit-to-stand Doppler radar images of the elderly can indeed reflect their gender to a certain extent.