Stationary Human Detection in Indoor Complex Environments via Millimeter-Wave Radar

Millimeter-wave (mmWave) radar has emerged as a critical technology for non-contact human sensing in security and healthcare applications. While traditional Doppler radar is limited to moving targets, micro-Doppler signatures enable the detection of stationary humans by capturing physiological signals such as respiration. However, in complex indoor environments, dynamic clutter sources (e.g., oscillating fans and swaying plants) generate periodic micro-Doppler signatures resembling human breathing, leading to frequent false alarms. To address this challenge, this article proposes a robust stationary human detection method based on sparse dictionary reconstruction. A Fourier transform-based dictionary is constructed to facilitate the sparse representation of micro-Doppler features. Leveraging the distinct sparsity characteristics of human respiration compared to environmental interference within this dictionary, the proposed method employs a Compressed Sensing (CS) framework. Specifically, the Orthogonal Matching Pursuit (OMP) algorithm is utilized to reconstruct signal components under strict sparsity constraints and physiological frequency priors. Target classification is then achieved by analyzing the reconstruction error, effectively isolating human signals from interference. Experimental results from real-world scenarios demonstrate that the proposed method achieves an interference suppression rate of 97% and a stationary human detection rate of 94%. Furthermore, in complex scenarios containing both targets and interference, the detection accuracy exceeds 86%. These results confirm that the proposed approach offers superior accuracy and robustness compared to existing methods.