Abstract: A posture recognition apparatus using radar according to the present invention includes a radar sensor module configured to process reflected signals reflected from objects positioned in a detection region and output a four-dimensional (4D) point cloud in units of frames, a tracking module configured to track the objects in the output 4D point cloud and reflect a position recognition result fed back from a posture recognition module to track the objects, and the posture recognition module configured to classify types of the objects for each track using a deep learning model based on pieces of output information about tracks, recognize a posture when the object is a person as a classification result of the track, and feed a posture recognition result of the corresponding track back to the tracking module.

Abstract: Provided is a radio detection and ranging (radar) point cloud-based posture determination system, which clusters a point cloud acquired using radar to generate a cluster, calculates predefined features of the cluster, and determines a posture of a target through a deep learning model that has trained the calculated features.

Abstract: In a point cloud of a 4-dimensional radar signal, Doppler information of each point is encoded with color information of that point. A 3-dimensional color point cloud is processed to recognize a shape of a target. A one-dimensional array feature vector generated by integration of feature maps extracted by processing 2-dimensional multi-view color point clouds with a convolution neural network (CNN) is processed by a recurrent neural network (RNN) to recognize the shape.