Hierarchical Multi-Domain Fusion Gait Recognition Method Based on FMCW Radar

Existing radar-based human gait recognition methods suffer from insufficient utilization of radar data, inadequate extraction of spectrogram feature information, and neglect of inter-domain complementary information. To address these issues, this paper proposes a Hierarchical Multi-Domain Gait Fusion Network (HMDGF-Net). Firstly, the FMCW radar is applied to collect the echo data of human gait movements. To fully utilize the radar information, the echo data is preprocessed to obtain three different radar domain spectrums. Subsequently, TFEM and DBFEM modules are designed based on the attributes of each spectrogram to perform feature extraction. Finally, based on the correlation between spectrograms, a hierarchical correlation fusion strategy is employed to fuse the extracted features, thereby enhancing the complementary information between different domains and achieving high-precision human gait recognition. Experimental results show that the proposed method achieves an average recognition accuracy of 95.28% for six types of gait movements, effectively improving the recognition performance compared to the state-of-the-art methods.