BiGRU Network for Human Activity Recognition in High Resolution Range Profile

The movements of the human body result in unique radar echo changes that can be exploited for classifying human activities. In this paper, we propose a bi-directional gated recurrent neural network (BiGRU) to classify human activities based on high resolution range profiles (HRRPs). The network is trained on a simulated dataset, which is generated by an ellipsoid-based human backscattering model using MOCAP dataset from Carnegie Mellon University. An accuracy of 97.5% on three activities is achieved with the BiGRU based classification approach.