Person Identification Based on Fine-Grained Micro-Doppler Signatures and UWB Radar

Person identification plays a vital role in preventing terrorist attacks and identifying illegal intrusions. In recent years, researchers have shown an increasing interest in radar-based person identification due to its advantages of robustness to the environment and protection of visual privacy. Person identification based on the micro-Doppler spectrogram is a typical fine-grained classification problem. Unlike the spectrograms of distinct actions, which have substantial variances, the micro-Doppler spectrograms of the same action from different people have only minor differences. In this article, we propose a person identification method based on a monostatic ultrawideband (UWB) radar system. Inspired by human visual habits, we establish a convolutional neural network (CNN) model based on the contrastive learning framework, together with the multiscale feature extraction and the image information interaction strategy, to solve the person identification problem. We perform the experiments on a dataset of 15 people, each performing six actions. The experimental results show that the proposed method has achieved an identification accuracy of 85.0% with the mixed-action model, and an average accuracy of 87.9% with the particular-action models. Finally, we discuss the effects of different slow-time window lengths and signal-to-noise ratios (SNRs).