MOVING TARGET CLASSIFICATION IN AUTOMOTIVE RADAR SYSTEMS USING TRANSPOSED CONVOLUTIONAL NETWORKS

Abstract

In this paper, we propose a deep neural network model for target classification in automotive radar system. In the proposed network, we introduce transposed convolutional network (TCNet) which applies transposed convolution operations. We discuss the properties of transposed convolution and show that TCNet can reduce the network size and improve the classification performance for the systems in which the signals are sparse and memory is restricted like our automotive radar systems. In our experiment, we show that the proposed network outperforms other popularly used dimensionality reduction approaches in terms of classification accuracy.