Research On Intelligent Recognition Method Of Underwater Acoustic Signal Modulation Type

As the competition between ocean powers becomes more and more fierce,the demand for data transmission between underwater equipment grows exponentially.In cooperative underwater acoustic communication,in order to improve communication efficiency,the sender usually uses adaptive modulation coding technology to select the optimal modulation mode for data transmission according to channel conditions.However,the complicated ocean noise often leads to the error of the handshake signal,which makes the receiver unable to accurately demodulate the signal,resulting in communication failure.In the non-cooperative underwater acoustic communication,the modulation mode adopted by the intercepted signal is usually not known in advance,and the receiver needs to demodulate data correctly in the absence of signal prior information.Therefore,how to accurately identify the modulation mode of the signal in real time is the key to the efficient and reliable underwater acoustic communication system.Automatic Modulation Recognition(AMR)technology can make the receiver automatically identify the Modulation mode of the signal when only receiving the underwater acoustic Modulation signal,so as to realize the correct demodulation,which is widely used in underwater acoustic communication with variable Modulation mode.The technology can be divided into Expert Feature Extraction-based AMR method and Automatic Feature Extraction-based AMR method.At present,there are the following problems: Firstly,the noise of underwater acoustic channel is complicated,accompanied by serious multipath interference and Doppler effect,which leads to serious damage to the modulation characteristics of underwater acoustic signal,thus affecting the recognition accuracy;Secondly,the lack of underwater acoustic signal samples makes it difficult to support data-driven deep learning algorithms.Thirdly,the existing deep learning architecture used in AMR cannot effectively extract the temporal and spatial features of underwater acoustic signals,which makes it difficult to achieve good recognition performance.Therefore,in view of the existing problems in the field of underwater acoustic signal modulation recognition,this paper proposes novel solutions from the two perspectives of Expert Feature Extraction-based AMR method and Automatic Feature Extraction-based AMR method respectively,and obtains excellent recognition results on the measured underwater acoustic signal data set.The main research work of this paper is as follows:(1)Optimizing Autoencoder(OAE)and Evaluation Enhanced K-nearest Neighbors(KNN)algorithms are proposed to solve the problem that signal characteristics are severely damaged and cannot be used in AMR under complex Marine environment.OAE can repair and optimize the distribution of damaged features by learning the mapping relationship between damaged features and ideal features,and significantly improve the distinguishing degree and availability of features.EEKNN makes up for the deficiency of traditional KNN algorithm through feature weighting and comprehensive evaluation.The combination of OAE and EEKNN can not only reduce redundant features,improve feature discrimination,but also effectively avoid misjudgment caused by abnormal samples,and achieve accurate and efficient automatic modulation recognition.(2)In view of the shortage of measured underwater acoustic signal samples,a data enhancement method is proposed for AMR.Through horizontal inversion,cross splicing and adding Gaussian white noise,the amount of underwater acoustic signal data is increased by 4 times without damaging the modulation characteristics of underwater acoustic signal,which solves the problem of insufficient data of AMR method based on depth learning and effectively improves the fitting performance of depth neural network.(3)Aiming at the problem that the existing deep learning architecture cannot effectively extract the temporal and spatial features of underwater acoustic signals,a Temporal Feature Extractor(One2Three Block)and a Double Squeeze-and-Excitation Networks(Dual Stream-SE Block)are proposed.Among them,One2 Three Block starts from the three micro scales of underwater acoustic signal,transforms the signal temporal features into three channel feature maps,and realizes the transformation from signal to image,so that the convolution module can extract spatial features.Dual Stream-SE Block optimizes the feature map while broadening the network width through the dual channel attention mechanism,and abstracts the feature map transformed by One2 Three Block from the perspective of spatial features,so as to realize the accurate recognition of modulation mode.(4)Based on the data set of the South China Sea experiment and the Yellow Sea experiment,the proposed algorithm is combined with Back Propagation Neural Network(BPNN),Support Vector Machines(SVM),Decision Trees(DT),Long-Short Term Memory(LSTM),Residual Network(Res Net),and Squeeze-and-Excitation Networks(SENet).The results show that the proposed method can achieve an average recognition accuracy of more than 99%,and the recognition rate is millisecond.The intelligent recognition of DSSS,2FSK,4FSK,2PSK,4PSK,16 QAM,64QAM and OFDM is successfully achieved,and the recognition performance is significantly better than the comparison algorithm.To sum up,this paper proposes two AMR methods based on expert feature extraction and artificial feature extraction,aiming at the main problems faced by automatic recognition technology of underwater acoustic signal modulation type in complex Marine environment.Aiming at the shortage of measured underwater acoustic signal,an efficient data enhancement method is proposed.The proposed algorithm achieves accurate and efficient intelligent recognition of underwater acoustic signal modulation type on experimental data sets in the South China Sea and the Yellow Sea,which verifies the superiority of the proposed algorithm.This research has greatly enriched the theoretical achievements in the field of automatic modulation recognition of underwater acoustic signals,provided theoretical support and practical guidance for improving the efficiency of underwater acoustic communication and the detection ability of underwater acoustic signals,and has great application value in both military and civil aspects,which is of great significance to enhance the strength of marine information technology in China.