利用新型卷积神经网络识别MPSK信号调制方式  

Modulation Recognition of MPSK Signals Based on Novel Convolutional Neural Network

作　　者：王爱丽[1] 张佳炜 姜开元 吴海滨[1] 岩堀祐之 WANG Ai-li;ZHANG Jia-wei;JIANG Kai-yuan;WU Hai-bin;Yuji Iwahori(School of Measurement-Control Technology and Communication Engineering, Harbin University of Science and Technology, Harbin 150080, China;Department of Computer Science, Chubu University, Aichi 487-8501, Japan)

机构地区：[1]哈尔滨理工大学测控技术与通信工程学院,哈尔滨150080 [2]中部大学计算机科学学院,日本爱知487-8501

出　　处：《哈尔滨理工大学学报》2021年第5期97-103,共7页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(61671190).

摘　　要：随着人工智能技术的快速发展,卷积神经网络越来越多地应用到通信信号调制识别领域。针对数字信号在低信噪比的识别准确率较低的问题,采用InceptionResNetV2网络与迁移适配相结合的方法,研究了一种调制识别模型,称之为InceptionResnetV2-TA,对MPSK信号的调制方式进行识别。结果表明,信噪比为3dB时,InceptionResnetV2-TA对BPSK的识别率达到99.33%,比次优模型InceptionResNetV2高出3%。对QPSK的识别率达到95.33%,比InceptionResNetV2高出2%。对8PSK的识别率达到86.33%,比InceptionResNetV2高出5%。综上,结合了迁移适配的InceptionResnetV2-TA,对BPSK、QPSK和8PSK在低信噪比的识别准确率高于其他对比方法。同时验证了这种调制识别模型的有效性。With the rapid development of artificial intelligence,convolutional neural network is more and more applied to the field of communication signal modulation recognition.Aiming at the problem of low recognition accuracy of digital signals at low SNR,a modulation recognition model named InceptionresnetV2-TA was studied by combining InceptionresnetV2 network with migration adaptation to identify the modulation mode of MPSK signals.The results show that when the SNR is 3dB,the recognition rate of InceptionresnetV2-TA for BPSK is 99.33%,which is 3%higher than that of the suboptimal model InceptionresnetV2.The recognition rate of QPSK is 95.33%,which is 2%higher than InceptionresnetV2.The recognition rate of 8PSK is 86.33%,which is 5%higher than that of Inceptionresnetv2.The above results indicate that InceptionresnetV2-TA combined with migration adaptation has higher identification accuracy of BPSK,QPSK and 8PSK at low SNR than other comparison methods.At the same time,the validity of the modulation recognition model is verified.

关 键 词：调制识别 卷积神经网络 迁移适配 InceptionResNetV2 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]