Research On Intra-class Modulation Recognition Technology Of Non-cooperative Underwater Acoustic Communication Signal

With more and more underwater acoustic sensor networks are deployed to perform tasks such as ocean data collection and ocean resource detection,ocean information security has attracted more and more attention.In order to protect ocean information security,it is necessary to develop underwater information countermeasure technology in which modulation recognition of non-cooperative underwater acoustic communication signal is an important research topic of underwater information countermeasure technology.Modulation recognition can be divided into two types:one is inter-class recognition of different types of modulation modes;the other is intra-class recognition under different types of modulation modes.Underwater acoustic channel is different from radio channel.It has complex multipath structure.Compared with the inter-class recognition of different types of modulation modes with typical recognition characteristics,the intra-class recognition of each type of modulation mode is more difficult.Therefore,it is challenging to realize highly reliable intra-class modulation recognition of non-cooperative underwater acoustic communication signals in underwater acoustic multipath channels.In this paper,the commonly used underwater acoustic communication signal intra-class modulation modes such as multiple frequency shift keying（MFSK）,orthogonal frequency division multiplexing（OFDM）and multiple phase shift keying（MPSK）are studied in underwater acoustic multipath channel.This paper solves the difficult problems involved in intra-class modulation recognition from three different points of view:extracting the robust features suitable for underwater acoustic multipath channel directly,time frequency representation（TFR）optimization and blind equalization of underwater acoustic multipath channel,so as to realize highly reliable intra-class modulation recognition of underwater acoustic communication signals and provide technical support for the establishment of a complete non-cooperative underwater acoustic communication signal modulation recognition system.In order to solve the problem of insufficient robustness of intra-class modulation recognition features extracted from MFSK spectrum in underwater acoustic multipath channel and high complexity of traditional machine learning classifiers,this paper studies non-cooperative underwater acoustic MFSK intra-class modulation recognition methods based on robust peak features combined with random forest（RF）,including 2FSK,4FSK and 8FSK.Firstly,the spectrum is optimized by taking the absolute value,peak enhancement,moving average filtering and Gaussian fitting.The optimized spectrum only has a large local maximum at the spectral peak,while the local maximums at other places are small or even tend to zero.Then,based on the characteristics of the optimized spectrum and the distribution of the local maximum at the spectral peak,the corresponding local maximum are extracted directly as the recognition features according to the different modulation modes in the intra-class.Finally,a RF classifier with low complexity and excellent performance is designed to realize efficient recognition.The simulation and experimental results show that:1.The peak feature extraction method can eliminate the influence of underwater acoustic multipath channel effectively and improve the robustness of peak features.2.RF classifier has higher recognition efficiency compared with the traditional support vector machine（SVM）and k-nearest neighbor（KNN）classifier.3.The proposed method of combining the robust peak feature with RF classifier can realize highly reliable MFSK intra-class modulation recognition in underwater acoustic multipath channel.In order to solve the problem that the lack of energy aggregation of TFR leads to the difficulty of carrier frequency extraction when using TFR to realize MFSK intra-class modulation recognition in underwater acoustic multipath channel,this paper studies the underwater acoustic MFSK intra-class modulation recognition method based on TFR optimization,including 2FSK,4FSK and 8FSK.Firstly,taking the inverse Fourier transform matrix as the observation matrix,combined with the characteristics of MFSK TFR,adding two constraints of element sparsity and row sparsity and the optimization model of sparse frequency reconstruction is established.Then,considering that the insufficient frequency resolution of TFR will lead to the inability to reconstruct the carrier frequency and serious energy divergence when the original MFSK carrier frequency is unknown under non-cooperative conditions,an inversion fitting method is further proposed which can improve the frequency resolution while keeping the TFR time resolution unchanged.Finally,the optimization model solution method of approximating the ₀l norm is proposed to obtain the optimized TFR.The simulation and experimental results show that:1.Compared with traditional methods,MFSK TFR reconstructed by sparse frequency has higher energy aggregation,which can reduce the impact of channel interference on TFR greatly.2.The inversion fitting method can solve the problem of insufficient frequency resolution when the MFSK carrier frequency is unknown,further reduce the energy divergence of TFR and improve the accuracy of carrier frequency reconstruction.3.The proposed TFR optimization method does not need any prior information,and is suitable for non-cooperative situations.In order to solve the problem of insufficient robustness of traditional OFDM intra-class modulation recognition features in underwater acoustic multipath channel,which leads to the failure of recognition methods,this paper studies the intra-class modulation recognition method of underwater acoustic OFDM based on blind equalization of subcarrier channel model,including BPSK,QPSK,8PSK and multiple quadrature amplitude modulation（MQAM）（M=16）.The proposed recognition method can be expressed as expectation maximization（EM）-block-maximum likelihood（ML）.Firstly,the channel characteristics of OFDM subcarriers are analyzed,and the block strategy for OFDM subcarrier mapping symbols is proposed according to the channel characteristics.Then,the EM algorithm is used to process the subcarrier mapping symbols in each block to estimate the channel fading coefficient and noise power under the block strategy.In order to improve the convergence speed of EM algorithm,K-means clustering algorithm is used to obtain the channel fading coefficient corresponding to the block mapping symbols,and the channel fading coefficient is taken as the initial value of EM algorithm,so that EM algorithm can converge quickly.Finally,the intra-class modulation recognition of OFDM is realized by ML classifier.The simulation and experimental results show that:1.K-means algorithm can accelerate the convergence speed of EM-Block algorithm.2.EM-Block algorithm can converge the chaotic and divergent subcarrier mapping symbols to the set subcarrier mapping symbols effectively and high blind equalization performance can be achieved,so as to solve the problem of unknown original subcarrier mapping symbols and ensure the effectiveness of ML classifier.3.The proposed EM-Block-ML method can realize highly reliable intra-class modulation recognition of OFDM in underwater acoustic multipath channel.In order to solve the problem of insufficient robustness of the traditional MPSK intra-class modulation recognition features in underwater acoustic multipath channel,which leads to the failure of traditional recognition methods,this paper studies the MPSK intra-class modulation recognition method based on blind equalization of single input multiple output（SIMO）channel model,including BPSK,QPSK and 8PSK.Firstly,the cross relationship between multi-channel can be derived based on SIMO channel model.Considering the sparse characteristics of underwater acoustic multipath channel,the blind channel estimation method based on orthogonal matching pursuit（OMP）and sparse Bayesian learning（SBL）is proposed.Then,the estimation results are used to equalize the received signal.Finally,MPSK intra class modulation recognition is realized from three aspects:high-order cumulant,clustering and ml classifier.The phase clustering is proposed to solve the problem of insufficient robustness of traditional clustering methods.The EM algorithm is further used to correct the received mapping symbol to the preset mapping symbol and realize the estimation of noise power.The simulation results show that:1.The blind channel estimation methods based on OMP and SBL shows better performance in channel impulse response（CIR）estimation.2.Compared with traditional clustering methods,phase clustering has higher robustness.3.The mapping symbol correction based on EM algorithm can effectively solve the problem of unknown original mapping symbols under non-cooperative conditions.4.The highly reliable MPSK intra-class modulation recognition is realized through a variety of recognition methods based on OMP and SBL blind channel estimation and equalization.