딥러닝을 이용한 동일 주파수 대역에 공존하는 통신 및 레이더 신호 분리

초록

When communication signals and radar signals coexist in the same frequency band, interference due to signal overlap inevitably occurs, resulting in degraded communication quality. Traditional frequency filtering methods are limited in performance when the frequencies completely overlap, which has led to the growing attention towards deep learning-based approaches. In this paper, U-Net and Conv-TasNet, deep learning models, are used to separate the overlapped communication and radar signals, and their performance is compared in terms of Bit Error Rate (BER). The experimental results show that, overall, the Conv-TasNet approach yields a lower BER than the U-Net approach. However, in environments with low Signal-to-Interference Ratio (SIR), U-Net shows a lower BER than Conv-TasNet.

키워드