基于TFAM-AVGNet的CR信号调制识别算法研究OA

Aiming at the problem of poor classification ability when facing complex signal types in the modulation recognition task of Cognitive Radio(CR)signals,an improved CR signal modulation recognition algorithm based on Temporal Fusion Attention Module(TFAM)and Adaptive Visibility Graph Neural Network(AVGNet)is proposed by combining the new graph deep learning theory.The problems such as overfitting of deep Graph Neural Network(GNN),gradient vanishing and simplistic feature fusion methods existing in the current AVGNet model are studied.By conducting data augmentation on the training set and introducing Residual Connections(RC)and TFAM,etc.,the gradient vanishing is alleviated and the feature reuse ability,as well as the training stability and convergence speed is improved.The experimental results show that the average recognition accuracy of the improved TFAM-AVGNet model is increased by more than 1.4%compared with the existing AVGNet model.