基于深度可分离卷积的自动调制识别OA

Automatic modulation recognition is an important part of communication pattern recognition,electronic reconnaissance,interfer-ence detection,and other fields.A lightweight network model based on depthwise separable convolution is proposed to address the issues of low accuracy in automatic modulation recognition under low signal-to-noise ratio conditions and a large number of network model parameters.This model mainly consists of depthwise separable convolution,RES-SCconv,and attention mechanism.Firstly,a basic feature extraction unit is constructed using depthwise separable convolution to effectively extract multi-scale features from time-frequency graph data;Second-ly,a lightweight attention mechanism is added to extract features from both spatial and channel dimensions,highlighting important feature in-formation;Finally,the RES-SCconv module is added to the network to further reduce feature redundancy in both spatial and channel dimen-sions,and suppress noise interference.The experimental results show that the proposed model achieves an average recognition accuracy of 90.68％on a dataset with a signal-to-noise ratio of-20～0 dB,and a recognition accuracy of 99.8％at a signal-to-noise ratio of 0 dB.Com-pared with the control model,the proposed model significantly improves the accuracy of automatic modulation recognition while maintaining lightweight.