Gender Classification of Elderly People using Doppler Radar Images based on Machine Learning

Nowadays, elderly care has become a social problem with the progress of the aging society. On the other hand, for privacy protection, the RGB images taken by camera are difficult to be used directly. For this reason, it is recommended to develop an elderly care system using Doppler radar images. In this paper, as the first step study of our elderly care system, we use Doppler radar images of elder people to classify the gender of them by machine learning and deep learning. Our research can be divided into three stages. Firstly, we use original Doppler radar images of walking and sit-to-stand process respectively for gender classification. Secondly, we do several processes on the sit-to-stand Doppler radar images for data augmentation and classify the gender again. Thirdly, we clip four parts of each sit-to-stand Doppler radar image and do gender classification. As a result, parts of experimental results show that classification accuracy achieve nearly 90%, which means that the sit-to-stand Doppler radar images tend to be influenced by the gender of elderly people.