Millimeter Wave Radar-Based Survivor Detection and Localization in Indoor Rescue Scenarios

This study introduces a survivor recognition and localization approach that leverages the temporal characteristics of complex-valued radar images and features obtained using millimeter-wave (MMW) sensors in assumed considerably optically blurred situations in indoor rescue scenarios, such as in firefighting. To address the challenges in localizing and detecting static survivors in low postures, we incorporate radar imaging and deep learning schemes to obtain temporal changes in complex-valued radar images to extract human-originated signs, such as respiration, heartbeat, and millimeter-scale trembling. For accurate, robust detection, this scheme introduces some promising feature vectors, such as spectral entropy and unwrapped phase sequences. Experimental validation using a 79 GHz MMW radar in an indoor environment shows that our proposed scheme discriminates a metallic cylinder object and a real human body with a recognition accuracy of over 90%.