Through-Wall Human Activity Classification Using Complex-Valued Convolutional Neural Network

Deep learning has attracted intensive attention in human activity classification based on the radar. Whereas, most methods use the images to classify the human activities, ignoring the phase information of the radar data. In this paper, the complex-valued convolutional neural network (Complex-valued CNN) is utilized to classify the human activity behind the wall. We developed several Complex-valued CNN models, which have the same structures as several classical convolutional neural network(CNN) models and use both the amplitude and phase information of the range profiles. Experiments on the real data validate the performance of the Complex-valued CNN models.