This dissertation proposes a vehicle occupant monitoring method using a mmWave wide-band planar radar array to monitor multiple occupants’ status. The radar array provides high range resolution with a wide field of view in both azimuth and elevation domain, making multiple occupant detection possible. Several methods are developed for posture detection, vital sign estimation, and classification of multiple occupants inside the vehicle cabin. Firstly, a mathematical model is proposed to describe the occupant reflection in the radio frequency environment. A signal processing pipeline is proposed based on the mathematical model. Next, a simulation framework is developed for the occupant’s posture detection and vital signs estimation. A reflection-based model is created to include both the size of each part of the human body and its reflection pattern for various sizes of the occupant simulation. A deep-learning-based method is then proposed based on the radar images reflected from the model for the occupant classification. This method utilizes the image information from the aligned camera as supervision to translate the radar point cloud to semantic segmentation masks. The designed network uses a sparse projected radar point cloud in 2D to generate occupants’ segmentation masks in different categories. The overall prediction accuracy of the designed method is acceptable and compatible with the accuracy of the camera-based image segmentation using the same network. For posture detection, a Keypoint-based model is proposed containing both posture and vital signs. Various sizes of the planer antennas are investigated. The optimal size of the antenna is selected to evaluate various human subjects in the simulation system. The results of the detection capability and accuracy are sufficient to distinguish the given sizes of the occupants as well as their liveness estimation in the simulation.