Multi-Input Deep Learning Based FMCW Radar Signal Classification

In autonomous driving vehicles, the emergency braking system uses lidar or radar sensors to recognize surrounding environment and prevent accidents. The conventional classifiers based on data using deep learning are single input structures range–Doppler maps micro-Doppler. Deep with a structure has limitations in improving classification performance. this paper, we propose multi-input classifier convolutional neural network (CNN) reduce amount of computation improve performance frequency modulated continuous wave (FMCW) radar. proposed is CNN-based distance Doppler map point cloud as multiple inputs. accuracy 85% 92%, respectively. It been improved 96% both maps.