【作者】 李楠；

【导师】 李秀坤；

【作者基本信息】 哈尔滨工程大学 ， 信号与信息处理， 2017， 博士

【摘要】 水下目标检测与识别是水声信号处理领域的核心内容,也是现代声呐系统与水声对抗领域的一个重要组成部分。不断提高的目标隐身技术水平及复杂多变的海洋环境,使得声呐接收信号的信噪比越来越低,低信噪比情况下的水下目标信号检测与识别成为现代水声信号处理的难点问题。本文借助典型非线性动力系统在微弱信号检测方面的特有优势,主要研究两方面内容:一是Duffing振子系统应用于水下目标信号检测的适用性问题,包括系统相态的定量判别方法及系统对不同分布特征噪声的免疫性研究;二是针对主、被动检测问题,研究与之匹配的水下弱目标信号的Duffing振子检测方法。研究Duffing振子对水下目标信号检测的适用性,需要解决两个关键问题。其一是为实现水下目标信号的检测与跟踪,需要一种能准确、快速且自动地对Duffing振子系统相态进行定量判别的方法,解决现有的定量判别方法存在计算数据量大、算法复杂及收敛速度慢等问题。通过对Duffing振子系统的非线性动力学行为研究,根据不同相态下Poincare映射的差异性,构造了一个可量化描述系统相态的度量参数,即Poincare映射特征函数(Poincare mapping characteristic function,PMCF),并提出了基于PMCF的系统相态定量判别方法。与其他定量算法对比分析表明,该方法能准确、快速且自动地对系统相态进行识别,无需人为干预。关键问题二是Duffing振子系统对服从不同分布特征噪声的免疫性分析。分别研究了Duffing振子系统对服从均值为零的高斯分布、混合高斯分布、对称Alpha稳定分布及K分布噪声的免疫情况。研究结果表明,Duffing振子对上述四种分布特征噪声均具有较强免疫性。研究被动探测中参数未知线谱的Duffing振子检测问题。当被动声呐探测舰船目标时,舰船辐射噪声中的线谱成分是舰船分类识别及航行状态监测的特征参量,通常情况下线谱频率及数量是未知的。针对现有未知频率的Duffing振子检测方法存在系统结构复杂,判别时间长,需要人为参与等问题,构造变参数单一Duffing振子检测模型,并提出了变频频率切片小波变换的Duffing振子检测方法。该方法通过设置合理的频率灵敏度参数自动地调整系统内置策动力频率值,利用频率切片小波变换提升系统抗噪性能,结合PMCF相态定量判决方法,变频搜索待测信号中的线谱分量。在实测数据的分析处理中,对不同观测时间内的数据采用本文方法进行检测,实现了线谱成分的检测跟踪与频率估计,并根据时间轨迹上的频率识别信息确定存在的稳定线谱成分。数据处理结果表明,该方法能够实现低信噪比下参数未知线谱的检测与频率估计。研究Duffing振子在水下主动目标探测中的应用,针对目标散射特性以及水声传播特性对声呐接收回波的影响,考虑水中沉底与掩埋目标探测时存在的混响干扰强、信混比低等问题,分别研究适用于发射信号为单频矩形(Continue Wave,CW)脉冲信号及线性调频信号(Linear frequency modulation,LFM)的Duffing振子检测方法。当发射CW脉冲信号时,多个声散射成分相干叠加使接收回波信号出现振幅起伏、衰落与相位改变的现象,而这些参数的变化严重影响了Duffing振子检测性能。针对该问题,提出了变系数短时支路互补的Duffing振子检测方法。为优化系统性能,降低系统漏检概率,提出了变系数PMCF相态定量判别方法。支路互补消除了相位参数变化使Duffing振子系统陷入检测盲区的问题;短时及变系数处理消除了因为振幅起伏、衰减而无法激励系统相变的问题。采用该方法对湖试数据进行处理,成功地实现了单目标及多目标信号的检测。在发射LFM信号时,目标回波为多个LFM信号线性叠加的形式,不符合Duffing振子的单频检测条件,对目标回波信号的解调频变换是首要问题。基于分数阶傅里叶变换(Fractional Fourier transform,FRFT)理论,提出了在最佳FRFT域上对LFM回波信号进行傅里叶逆变换的解调频方法。根据亮点模型,理论推导了多亮点LFM回波信号解调频后的信号形式。解调频后的单频信号数量与亮点数量一致,频率与亮点时延及发射信号参数有关,且为未知参量。在此基础上,提出了一种基于FRFT域谱峭度的变频Duffing振子检测方法。该方法根据最佳FRFT域上谱峭度值确定LFM回波信号存在区间,将该区间信号解调频后送入变频Duffing振子系统检测,自主地搜索单频信号信息,实现了低信混比下LFM目标回波信号的检测及亮点数量的估计。湖试沉底及海试掩埋实验数据处理结果,验证了方法的有效性及可行性。上述研究成果证明了Duffing振子对低信噪比水下目标信号具有良好的检测性能,拓展了水声信号处理的研究思路及途径。更多还原

【Abstract】 Underwater target detection and identification is the key focus area in the field of underwater acoustic signal processing,as well as an important part of modern sonar system and underwater acoustic countermeasure field.The continuous improvement of stealth technology and the complex and changeable ocean environment have caused the signal to noise ratio getting lower.Consequently,underwater target signal detection and identification in the condition of low signal to noise ratio(SNR)is becoming a difficult problem in modern underwater acoustic signal processing.By utilizing the advantage of typical nonlinear dynamic system in weak signal detection,studies are carried out in two aspects: one is the applicability research of the Duffing oscillator system in underwater target signal detection,which consists of researching the method of the system phase quantitative discrimination and studying the immunity of the system to the noise with different distribution characteristics.The other is,aiming at active and passive detection,to study the matching method of using the Duffing oscillator in detecting underwater weak target signal.Two key problems need to be studied in the application of Duffing oscillator to underwater target detection.Firstly,in order to successfully detect and track the signal of underwater target,it is necessary to find out a quantitative method to accurately,quickly and automatically discriminate phase state of the Duffing oscillator system,wihch solve the problems with the current quantitative discrimination method,such as large amounts of data processing,complex calculation steps and slow convergence speed.Based on the study on the nonlinear dynamic behavior of the Duffing oscillator system,according to the difference of Poincare mapping in different phase states,a parameter which can quantify the phase state of the system,i.e.,the Poincare mapping characteristic function(PMCF),has been constructed.A method for quantitative determination of system phase states based on PMCF has been proposed in this paper.Comparing with other quantifying methods,it can accurately,quickly and automatically identify the phase state of the system without human intervention.Secondly,it is also critical to analyze the immunity of the Duffing oscillator system to underwater noises with different distribution characteristics.In this paper,the sensitivity of Duffing oscillator system to noises with the characteristics of the Gauss distribution,the mixed Gauss distribution,the symmetric Alpha stable distribution and the K distribution has been studied.The numerical simulation results show that Duffing oscillator has strong immunity to the noises presented with the above four distribution characteristics.Passive detection of the line spectrum with unknown parameters by using Duffing oscillator has been studied.The line spectrum of ship-radiated noise is an important characteristic parameter of ship classification and navigation monitoring in the circumstances of sonar detecting ship targets passively.Usually,the frequency and quantity of line spectrum are unknown.Aiming at resolving the problems with the current Duffing oscillator system in detecting unknown frequency signal,such as complex structure,lengthy distinguishment,and the need for human participation,a single Duffing oscillator detection model with variable parameters is constructed,and a Duffing oscillator detection method with variable frequency scanning based on frequency slice wavelet transform(VFFSWT)has been proposed.By setting a parameter of the frequency sensitivity,the frequency of the system driving force is adjusted automatically,and combining with the PMCF method,components of the line spectrum are automatically detected.In this method,the anti-noise performance of the system can be improved by using the frequency slice wavelet transformation.In the processing of the measured data,the VFFSWT-Duffing method is used to detect data at different observation times in order to detect and trace the line spectrum and estimate frequency.Stable line spectrum is identified according to the frequency information on the time track.The processing results show that the method is feasible and effective.The application of Duffing oscillator in active underwater target detection has been studied.Focusing on the influence of target scatting and acoustic propagation characteristics on the characteristics of sonar echo,and considering the problems with underwater bottom and buried targets detection,such as strong reverberation interference,etc.,the weak signal detection methods for the continue wave pulse signal(CW)signal and linear frequency modulation signal(LFM)are studied.When the CW pulse signal is transmitted,the multiple scattering components are coherently superposed in the target echo,which resulting in the received CW echo signal presenting the phenomenon of amplitude fluctuating,fading and phase changing.Consequently,the detection performance of the Duffing oscillator is seriously impacted by these changing parameters.To solve this problem,a new method of the Duffing oscillator with variable coefficients and short-time branch complementation is proposed.In order to optimize the system performance and reduce the system failure rate,the variable coefficient PMCF method is proposed.The strategy of branch complementation solved the phase changing driving the system into the detection blind area.The short time and variable coefficient processing reduces false dismissal probability caused by amplitude fluctuation and fading of the received signal.This method is used to process the data collected in lakes,and presented a great success in single target and multi-target signal detection.When the transmitted signal is LFM signal,the target echo is a linear superposition of multiple LFM signals,which does not meet the single frequency detection condition of the Duffing oscillator.The frequency demodulation of the target echo signal is the main issue.Based on the theory of fractional Fourier transform(FRFT),in the optimal FRFT domain,infinite LFM signal exhibits spectral characteristics of impulse function,and its Fourier transform is a single frequency signal.The above process can achieve demodulation of LFM echo signal.Based on the highlight model,the signal form of the multi-highlight LFM echo signal demodulation is deduced.After demodulation,the number of single frequency components is the same as the number of highlights,and the frequency is related to the delay parameters of the highlight and the parameters of the transmitted signal.On this basis,a Duffing oscillator detection method based spectrum kurtosis in FRFT domain is proposed.By using this method,the existence interval of target echo is determined based on the spectral kurtosis value of optimal FRFT domain.The signal is demodulated and sent to the Duffing oscillator system.The single frequency components are searched automatically,which can realize the detection of LFM target echo signal and the estimation of the number of highligts.The validity and feasibility of the method is verified by the results of processing data collected in the lake trial and ocean trial.The above research results prove that the Duffing oscillator has good detection performance for underwater target signal with low signal to noise ratio.The research ideas and approaches of underwater acoustic signal processing are developed.更多还原