Millimeter-wave human inspection techniques have drawn much attention in inspection sites due to their harmlessness, penetrability, and contactless imaging. Existing approaches utilize intensity images to detect concealed objects. However, the intensity features of small objects are often weak and limited. The millimeter-wave depth image provides complementary information, but it contains noisy interference. To this end, this paper proposes a trusted depth knowledge transfer network to integrate useful and discriminative depth knowledge, enhancing the discrimination for concealed objects. Specifically, the proposed method uses a Teacher-Student architecture that treats the depth image-based detection network as the teacher and the intensity image as the student. To distill the reliable and discriminative depth knowledge into the student network, a confidence-aware trusted knowledge transmission module is designed. Experiments on the AMMW-3D human inspection dataset show that our proposed method effectively excavates complementary discriminative information and filters out depth information with side effects, improving detection performances.