Radar Object Classification Using Neural Networks in Urban Automotive Scenarios

This paper proposes a method to solve the problem of object classification for automotive radar systems. This is done using neural networks and knowing attributes about the objects such as their trajectory, acceleration, velocity, velocity and acceleration standard deviation, object length and width and other parameters related to object's movement. The method proposed is validated by experiments, obtaining in process of validation an overall accuracy over 99%: for the pedestrian class an accuracy of 99.88%, for the cyclist class an accuracy of 99.23% and for the car class 99.82% accuracy.