개선된 손실 함수를 이용한 딥러닝 기반 자동 변조 분류

초록

Automatic Modulation Classification(AMC) is one of the key technologies of modern wireless communication which plays an important role in various cooperative and non-cooperative contexts. Recently, many studies on Deep Learning(DL)-based AMC have been reported. This paper proposes a method for improving classification performance by modifying the loss function of the DL model for AMC and analyzes its classification performance. The proposed method improves the conventional softmax loss function to adjust the probability distribution over the modulation schemes closer to the desired probability distribution. Through computer simulations, we verify that by applying the loss function revised with the proposed method, it is possible to improve performance in terms of classification accuracy for various DL models than conventional ones.

키워드