基于自适应空间通道收缩网络的自动调制识别算法OA

In view of the insufficient feature learning in automatic modulation recognition(AMR)algorithm at low signal-to-noise ratio(SNR),this paper proposes an AMR algorithm based on adaptive space channel shrinkage network.The algorithm is mainly composed of adaptive space channel shrinkage(ASCS)module,multi-scale convolution(MC)module,residual module and multi-head attention(MHA)module.The ASCS module is used to extract features in the spatial and channel dimensions,and the improved soft threshold function is used to shrink the features to reduce noise features and retain useful features,so as to improve the feature processing ability of the network.Comprehensive analysis of experimental results show that the improved soft threshold function can process the features better.The average recognition accuracy of the proposed AMR algorithm is 62.94%on the public dataset RML2016.10a,and 64.79%on RML2016.10b.In comparison with the existing AMR algorithm,the proposed algorithm can achieve higher accuracy.In conclusion,it provides an effective method for feature learning at low SNR.