Estimation of Number of Targets Based on CNN Classifier for OFDM Radar Systems

Authors

  • Jae-Woong Choi, Jeong-Eun Oh, A-Min Jo, Eui-Rim Jeong

DOI:

https://doi.org/10.17762/msea.v71i3.194

Abstract

This paper proposes a new convolutional neural network (CNN) for estimating the number of targets in orthogonal frequency division multiplexing (OFDM) radar systems. The transmitter of OFDM radar system receives the signal reflected on the target. Two-dimensional periodogram is obtained via 2D fast Fourier transform (FFT) from the reflected signal after removing the modulation effect. Two-dimensional periodogram is an input signal for the CNN classifier. CNN classifier estimates the number of targets. We propose two types of CNN models. One is 7-layer CNN model and the other is 4-layer CNN model. The maximum number of targets was set to 2, 4 and 8. According to the simulation, the accuracy is degraded as the number of targets increased. Comparing the proposed 7-layer model and 4-layer model, the detection accuracy of 7-layer is about 23.9% better than 4-layer. However, 4-layer has much lower complexity than 7-layer.

Downloads

Published

2022-06-09

How to Cite

Jae-Woong Choi, Jeong-Eun Oh, A-Min Jo, Eui-Rim Jeong. (2022). Estimation of Number of Targets Based on CNN Classifier for OFDM Radar Systems. Mathematical Statistician and Engineering Applications, 71(3), 555 –. https://doi.org/10.17762/msea.v71i3.194

Issue

Section

Articles