The Research And Implementation Of Signal Modul Ation Recognition Technology Based On Deep Learning

Signal modulation recognition technology has been a hot research topic during the long development process of wireless communication.With the development of 5G networks and IoT technologies,the wireless communication environment is becoming more complex,and the number of signal types is gradually increasing.Traditional signal modulation recognition methods have been difficult to meet contemporary communication requirements.As the most popular technology in the era of artificial intelligence,deep learning has made fruitful achievements in the fields of computer vision and natural language processing.Convolutional neural networks can solve problems such as pattern recognition,feature fitting and multi-dimensional classification.Therefore,how to apply deep learning technology into wireless communication fields to overcome and break through the shortcomings of traditional modulation recognition technology is the focus of this paper.This paper regards the signal modulation recognition task as a classification problem in deep learning,and designs a basic signal automatic modulation recognition neural network(AMRNet).In addition,considering the time complexity of convolutional neural networks,this paper proposes to replace the standard convolution way with a lightweight network structure,which improves the processing efficiency of the communication systems without losing the accuracy of recognition.Secondly,based on the analysis of the communication signal data,this paper proposes to use the attention mechanism to extract the features of the communication signal from the two dimensions of time and space to improve the recognition accuracy.In the practical application of wireless communication systems,the communication quality is greatly affected by noise,so this paper further proposes two algorithms for signal modulation recognition under low signal-to-noise ratio conditions.This paper proposes a special focus loss function to increase the network’s attention to low SNR signals,thereby improving the accuracy of signal recognition.In addition,this paper proposes to use Generative Adversarial Networks(GAN)to denoise low SNR signals,thereby significantly improving the accuracy of signal modulation recognition.Finally,based on USRP N210 hardware platform,GNU Radio software platform and PyTorch deep learning platform,this paper builds a wireless communication system for automatic recognition of signal modulation,which proves the practicability of our proposed method.