Function modules: First, voice control, offline voice recognition, response <0.3 seconds, supporting dialects and multiple commands; Second, gesture recognition, AI vision recognizes waving, fist clenching and up-down gestures to control door and floor selection; Third, 3D face, infrared binocular camera, living detection, recognition rate 95%+ with masks and hats; Fourth, multi-modal fusion, automatically switching to optimal mode, stable operation in low light and backlight; Fifth, privacy protection, local algorithm processing, no face data uploading, in compliance with Data Security Law.